ufs_vnops.c revision 67893
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 67893 2000-10-29 16:06:56Z phk $ 40 */ 41 42#include "opt_quota.h" 43#include "opt_suiddir.h" 44 45#include <sys/param.h> 46#include <sys/systm.h> 47#include <sys/namei.h> 48#include <sys/kernel.h> 49#include <sys/fcntl.h> 50#include <sys/stat.h> 51#include <sys/bio.h> 52#include <sys/buf.h> 53#include <sys/mount.h> 54#include <sys/unistd.h> 55#include <sys/vnode.h> 56#include <sys/malloc.h> 57#include <sys/dirent.h> 58#include <sys/lockf.h> 59#include <sys/event.h> 60#include <sys/conf.h> 61 62#include <machine/mutex.h> 63 64#include <vm/vm.h> 65#include <vm/vm_extern.h> 66 67#include <miscfs/fifofs/fifo.h> 68 69#include <ufs/ufs/extattr.h> 70#include <ufs/ufs/quota.h> 71#include <ufs/ufs/inode.h> 72#include <ufs/ufs/dir.h> 73#include <ufs/ufs/ufsmount.h> 74#include <ufs/ufs/ufs_extern.h> 75 76static int ufs_access __P((struct vop_access_args *)); 77static int ufs_advlock __P((struct vop_advlock_args *)); 78static int ufs_chmod __P((struct vnode *, int, struct ucred *, struct proc *)); 79static int ufs_chown __P((struct vnode *, uid_t, gid_t, struct ucred *, struct proc *)); 80static int ufs_close __P((struct vop_close_args *)); 81static int ufs_create __P((struct vop_create_args *)); 82static int ufs_getattr __P((struct vop_getattr_args *)); 83static int ufs_link __P((struct vop_link_args *)); 84static int ufs_makeinode __P((int mode, struct vnode *, struct vnode **, struct componentname *)); 85static int ufs_missingop __P((struct vop_generic_args *ap)); 86static int ufs_mkdir __P((struct vop_mkdir_args *)); 87static int ufs_mknod __P((struct vop_mknod_args *)); 88static int ufs_mmap __P((struct vop_mmap_args *)); 89static int ufs_open __P((struct vop_open_args *)); 90static int ufs_pathconf __P((struct vop_pathconf_args *)); 91static int ufs_print __P((struct vop_print_args *)); 92static int ufs_readdir __P((struct vop_readdir_args *)); 93static int ufs_readlink __P((struct vop_readlink_args *)); 94static int ufs_remove __P((struct vop_remove_args *)); 95static int ufs_rename __P((struct vop_rename_args *)); 96static int ufs_rmdir __P((struct vop_rmdir_args *)); 97static int ufs_setattr __P((struct vop_setattr_args *)); 98static int ufs_strategy __P((struct vop_strategy_args *)); 99static int ufs_symlink __P((struct vop_symlink_args *)); 100static int ufs_whiteout __P((struct vop_whiteout_args *)); 101static int ufsfifo_close __P((struct vop_close_args *)); 102static int ufsfifo_read __P((struct vop_read_args *)); 103static int ufsfifo_write __P((struct vop_write_args *)); 104static int ufsspec_close __P((struct vop_close_args *)); 105static int ufsspec_read __P((struct vop_read_args *)); 106static int ufsspec_write __P((struct vop_write_args *)); 107 108union _qcvt { 109 int64_t qcvt; 110 int32_t val[2]; 111}; 112#define SETHIGH(q, h) { \ 113 union _qcvt tmp; \ 114 tmp.qcvt = (q); \ 115 tmp.val[_QUAD_HIGHWORD] = (h); \ 116 (q) = tmp.qcvt; \ 117} 118#define SETLOW(q, l) { \ 119 union _qcvt tmp; \ 120 tmp.qcvt = (q); \ 121 tmp.val[_QUAD_LOWWORD] = (l); \ 122 (q) = tmp.qcvt; \ 123} 124#define VN_KNOTE(vp, b) \ 125 KNOTE(&vp->v_pollinfo.vpi_selinfo.si_note, (b)) 126 127/* 128 * A virgin directory (no blushing please). 129 */ 130static struct dirtemplate mastertemplate = { 131 0, 12, DT_DIR, 1, ".", 132 0, DIRBLKSIZ - 12, DT_DIR, 2, ".." 133}; 134static struct odirtemplate omastertemplate = { 135 0, 12, 1, ".", 136 0, DIRBLKSIZ - 12, 2, ".." 137}; 138 139void 140ufs_itimes(vp) 141 struct vnode *vp; 142{ 143 struct inode *ip; 144 struct timespec ts; 145 146 ip = VTOI(vp); 147 if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE)) == 0) 148 return; 149 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) { 150 vfs_timestamp(&ts); 151 if ((vp->v_type == VBLK || vp->v_type == VCHR) && 152 !DOINGSOFTDEP(vp)) 153 ip->i_flag |= IN_LAZYMOD; 154 else 155 ip->i_flag |= IN_MODIFIED; 156 if (ip->i_flag & IN_ACCESS) { 157 ip->i_atime = ts.tv_sec; 158 ip->i_atimensec = ts.tv_nsec; 159 } 160 if (ip->i_flag & IN_UPDATE) { 161 ip->i_mtime = ts.tv_sec; 162 ip->i_mtimensec = ts.tv_nsec; 163 ip->i_modrev++; 164 } 165 if (ip->i_flag & IN_CHANGE) { 166 ip->i_ctime = ts.tv_sec; 167 ip->i_ctimensec = ts.tv_nsec; 168 } 169 } 170 ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE); 171} 172 173/* 174 * Create a regular file 175 */ 176int 177ufs_create(ap) 178 struct vop_create_args /* { 179 struct vnode *a_dvp; 180 struct vnode **a_vpp; 181 struct componentname *a_cnp; 182 struct vattr *a_vap; 183 } */ *ap; 184{ 185 int error; 186 187 error = 188 ufs_makeinode(MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode), 189 ap->a_dvp, ap->a_vpp, ap->a_cnp); 190 if (error) 191 return (error); 192 VN_KNOTE(ap->a_dvp, NOTE_WRITE); 193 return (0); 194} 195 196/* 197 * Mknod vnode call 198 */ 199/* ARGSUSED */ 200int 201ufs_mknod(ap) 202 struct vop_mknod_args /* { 203 struct vnode *a_dvp; 204 struct vnode **a_vpp; 205 struct componentname *a_cnp; 206 struct vattr *a_vap; 207 } */ *ap; 208{ 209 struct vattr *vap = ap->a_vap; 210 struct vnode **vpp = ap->a_vpp; 211 struct inode *ip; 212 int error; 213 214 error = ufs_makeinode(MAKEIMODE(vap->va_type, vap->va_mode), 215 ap->a_dvp, vpp, ap->a_cnp); 216 if (error) 217 return (error); 218 VN_KNOTE(ap->a_dvp, NOTE_WRITE); 219 ip = VTOI(*vpp); 220 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE; 221 if (vap->va_rdev != VNOVAL) { 222 /* 223 * Want to be able to use this to make badblock 224 * inodes, so don't truncate the dev number. 225 */ 226 ip->i_rdev = vap->va_rdev; 227 } 228 /* 229 * Remove inode, then reload it through VFS_VGET so it is 230 * checked to see if it is an alias of an existing entry in 231 * the inode cache. 232 */ 233 vput(*vpp); 234 (*vpp)->v_type = VNON; 235 vgone(*vpp); 236 error = VFS_VGET(ap->a_dvp->v_mount, ip->i_ino, vpp); 237 if (error) { 238 *vpp = NULL; 239 return (error); 240 } 241 return (0); 242} 243 244/* 245 * Open called. 246 * 247 * Nothing to do. 248 */ 249/* ARGSUSED */ 250int 251ufs_open(ap) 252 struct vop_open_args /* { 253 struct vnode *a_vp; 254 int a_mode; 255 struct ucred *a_cred; 256 struct proc *a_p; 257 } */ *ap; 258{ 259 260 /* 261 * Files marked append-only must be opened for appending. 262 */ 263 if ((VTOI(ap->a_vp)->i_flags & APPEND) && 264 (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE) 265 return (EPERM); 266 return (0); 267} 268 269/* 270 * Close called. 271 * 272 * Update the times on the inode. 273 */ 274/* ARGSUSED */ 275int 276ufs_close(ap) 277 struct vop_close_args /* { 278 struct vnode *a_vp; 279 int a_fflag; 280 struct ucred *a_cred; 281 struct proc *a_p; 282 } */ *ap; 283{ 284 register struct vnode *vp = ap->a_vp; 285 286 mtx_enter(&vp->v_interlock, MTX_DEF); 287 if (vp->v_usecount > 1) 288 ufs_itimes(vp); 289 mtx_exit(&vp->v_interlock, MTX_DEF); 290 return (0); 291} 292 293int 294ufs_access(ap) 295 struct vop_access_args /* { 296 struct vnode *a_vp; 297 int a_mode; 298 struct ucred *a_cred; 299 struct proc *a_p; 300 } */ *ap; 301{ 302 struct vnode *vp = ap->a_vp; 303 struct inode *ip = VTOI(vp); 304 mode_t mode = ap->a_mode; 305#ifdef QUOTA 306 int error; 307#endif 308 309 /* 310 * Disallow write attempts on read-only file systems; 311 * unless the file is a socket, fifo, or a block or 312 * character device resident on the file system. 313 */ 314 if (mode & VWRITE) { 315 switch (vp->v_type) { 316 case VDIR: 317 case VLNK: 318 case VREG: 319 if (vp->v_mount->mnt_flag & MNT_RDONLY) 320 return (EROFS); 321#ifdef QUOTA 322 if ((error = getinoquota(ip)) != 0) 323 return (error); 324#endif 325 break; 326 default: 327 break; 328 } 329 } 330 331 /* If immutable bit set, nobody gets to write it. */ 332 if ((mode & VWRITE) && (ip->i_flags & (IMMUTABLE | SF_SNAPSHOT))) 333 return (EPERM); 334 335 return (vaccess(vp->v_type, ip->i_mode, ip->i_uid, ip->i_gid, 336 ap->a_mode, ap->a_cred, NULL)); 337} 338 339/* ARGSUSED */ 340int 341ufs_getattr(ap) 342 struct vop_getattr_args /* { 343 struct vnode *a_vp; 344 struct vattr *a_vap; 345 struct ucred *a_cred; 346 struct proc *a_p; 347 } */ *ap; 348{ 349 register struct vnode *vp = ap->a_vp; 350 register struct inode *ip = VTOI(vp); 351 register struct vattr *vap = ap->a_vap; 352 353 ufs_itimes(vp); 354 /* 355 * Copy from inode table 356 */ 357 vap->va_fsid = dev2udev(ip->i_dev); 358 vap->va_fileid = ip->i_number; 359 vap->va_mode = ip->i_mode & ~IFMT; 360 vap->va_nlink = VFSTOUFS(vp->v_mount)->um_i_effnlink_valid ? 361 ip->i_effnlink : ip->i_nlink; 362 vap->va_uid = ip->i_uid; 363 vap->va_gid = ip->i_gid; 364 vap->va_rdev = ip->i_rdev; 365 vap->va_size = ip->i_din.di_size; 366 vap->va_atime.tv_sec = ip->i_atime; 367 vap->va_atime.tv_nsec = ip->i_atimensec; 368 vap->va_mtime.tv_sec = ip->i_mtime; 369 vap->va_mtime.tv_nsec = ip->i_mtimensec; 370 vap->va_ctime.tv_sec = ip->i_ctime; 371 vap->va_ctime.tv_nsec = ip->i_ctimensec; 372 vap->va_flags = ip->i_flags; 373 vap->va_gen = ip->i_gen; 374 vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize; 375 vap->va_bytes = dbtob((u_quad_t)ip->i_blocks); 376 vap->va_type = IFTOVT(ip->i_mode); 377 vap->va_filerev = ip->i_modrev; 378 return (0); 379} 380 381/* 382 * Set attribute vnode op. called from several syscalls 383 */ 384int 385ufs_setattr(ap) 386 struct vop_setattr_args /* { 387 struct vnode *a_vp; 388 struct vattr *a_vap; 389 struct ucred *a_cred; 390 struct proc *a_p; 391 } */ *ap; 392{ 393 struct vattr *vap = ap->a_vap; 394 struct vnode *vp = ap->a_vp; 395 struct inode *ip = VTOI(vp); 396 struct ucred *cred = ap->a_cred; 397 struct proc *p = ap->a_p; 398 int error; 399 400 /* 401 * Check for unsettable attributes. 402 */ 403 if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) || 404 (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) || 405 (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) || 406 ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) { 407 return (EINVAL); 408 } 409 if (vap->va_flags != VNOVAL) { 410 if (vp->v_mount->mnt_flag & MNT_RDONLY) 411 return (EROFS); 412 /* 413 * Callers may only modify the file flags on objects they 414 * have VADMIN rights for. 415 */ 416 if ((error = VOP_ACCESS(vp, VADMIN, cred, p))) 417 return (error); 418 /* 419 * Unprivileged processes and privileged processes in 420 * jail() are not permitted to set system flags. 421 * Privileged processes not in jail() may only set system 422 * flags if the securelevel <= 0. 423 */ 424 if (!suser_xxx(cred, NULL, 0)) { 425 if ((ip->i_flags 426 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND)) && 427 securelevel > 0) 428 return (EPERM); 429 /* Snapshot flag cannot be set or cleared */ 430 if (((vap->va_flags & SF_SNAPSHOT) != 0 && 431 (ip->i_flags & SF_SNAPSHOT) == 0) || 432 ((vap->va_flags & SF_SNAPSHOT) == 0 && 433 (ip->i_flags & SF_SNAPSHOT) != 0)) 434 return (EPERM); 435 ip->i_flags = vap->va_flags; 436 } else { 437 if (ip->i_flags 438 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND) || 439 (vap->va_flags & UF_SETTABLE) != vap->va_flags) 440 return (EPERM); 441 ip->i_flags &= SF_SETTABLE; 442 ip->i_flags |= (vap->va_flags & UF_SETTABLE); 443 } 444 ip->i_flag |= IN_CHANGE; 445 if (vap->va_flags & (IMMUTABLE | APPEND)) 446 return (0); 447 } 448 if (ip->i_flags & (IMMUTABLE | APPEND)) 449 return (EPERM); 450 /* 451 * Go through the fields and update iff not VNOVAL. 452 */ 453 if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) { 454 if (vp->v_mount->mnt_flag & MNT_RDONLY) 455 return (EROFS); 456 if ((error = ufs_chown(vp, vap->va_uid, vap->va_gid, cred, 457 p)) != 0) 458 return (error); 459 } 460 if (vap->va_size != VNOVAL) { 461 /* 462 * Disallow write attempts on read-only file systems; 463 * unless the file is a socket, fifo, or a block or 464 * character device resident on the file system. 465 */ 466 switch (vp->v_type) { 467 case VDIR: 468 return (EISDIR); 469 case VLNK: 470 case VREG: 471 if (vp->v_mount->mnt_flag & MNT_RDONLY) 472 return (EROFS); 473 if ((ip->i_flags & SF_SNAPSHOT) != 0) 474 return (EPERM); 475 break; 476 default: 477 break; 478 } 479 if ((error = UFS_TRUNCATE(vp, vap->va_size, 0, cred, p)) != 0) 480 return (error); 481 } 482 if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) { 483 if (vp->v_mount->mnt_flag & MNT_RDONLY) 484 return (EROFS); 485 if ((ip->i_flags & SF_SNAPSHOT) != 0) 486 return (EPERM); 487 /* 488 * From utimes(2): 489 * If times is NULL, ... The caller must be the owner of 490 * the file, have permission to write the file, or be the 491 * super-user. 492 * If times is non-NULL, ... The caller must be the owner of 493 * the file or be the super-user. 494 */ 495 if ((error = VOP_ACCESS(vp, VADMIN, cred, p)) && 496 ((vap->va_vaflags & VA_UTIMES_NULL) == 0 || 497 (error = VOP_ACCESS(vp, VWRITE, cred, p)))) 498 return (error); 499 if (vap->va_atime.tv_sec != VNOVAL) 500 ip->i_flag |= IN_ACCESS; 501 if (vap->va_mtime.tv_sec != VNOVAL) 502 ip->i_flag |= IN_CHANGE | IN_UPDATE; 503 ufs_itimes(vp); 504 if (vap->va_atime.tv_sec != VNOVAL) { 505 ip->i_atime = vap->va_atime.tv_sec; 506 ip->i_atimensec = vap->va_atime.tv_nsec; 507 } 508 if (vap->va_mtime.tv_sec != VNOVAL) { 509 ip->i_mtime = vap->va_mtime.tv_sec; 510 ip->i_mtimensec = vap->va_mtime.tv_nsec; 511 } 512 error = UFS_UPDATE(vp, 0); 513 if (error) 514 return (error); 515 } 516 error = 0; 517 if (vap->va_mode != (mode_t)VNOVAL) { 518 if (vp->v_mount->mnt_flag & MNT_RDONLY) 519 return (EROFS); 520 if ((ip->i_flags & SF_SNAPSHOT) != 0 && (vap->va_mode & 521 (S_IXUSR | S_IWUSR | S_IXGRP | S_IWGRP | S_IXOTH | S_IWOTH))) 522 return (EPERM); 523 error = ufs_chmod(vp, (int)vap->va_mode, cred, p); 524 } 525 VN_KNOTE(vp, NOTE_ATTRIB); 526 return (error); 527} 528 529/* 530 * Change the mode on a file. 531 * Inode must be locked before calling. 532 */ 533static int 534ufs_chmod(vp, mode, cred, p) 535 register struct vnode *vp; 536 register int mode; 537 register struct ucred *cred; 538 struct proc *p; 539{ 540 register struct inode *ip = VTOI(vp); 541 int error; 542 543 /* 544 * To modify the permissions on a file, must possess VADMIN 545 * for that file. 546 */ 547 if ((error = VOP_ACCESS(vp, VADMIN, cred, p))) 548 return (error); 549 /* 550 * Privileged processes may set the sticky bit on non-directories, 551 * as well as set the setgid bit on a file with a group that the 552 * process is not a member of. 553 */ 554 if (suser_xxx(cred, NULL, PRISON_ROOT)) { 555 if (vp->v_type != VDIR && (mode & S_ISTXT)) 556 return (EFTYPE); 557 if (!groupmember(ip->i_gid, cred) && (mode & ISGID)) 558 return (EPERM); 559 } 560 ip->i_mode &= ~ALLPERMS; 561 ip->i_mode |= (mode & ALLPERMS); 562 ip->i_flag |= IN_CHANGE; 563 return (0); 564} 565 566/* 567 * Perform chown operation on inode ip; 568 * inode must be locked prior to call. 569 */ 570static int 571ufs_chown(vp, uid, gid, cred, p) 572 register struct vnode *vp; 573 uid_t uid; 574 gid_t gid; 575 struct ucred *cred; 576 struct proc *p; 577{ 578 register struct inode *ip = VTOI(vp); 579 uid_t ouid; 580 gid_t ogid; 581 int error = 0; 582#ifdef QUOTA 583 register int i; 584 long change; 585#endif 586 587 if (uid == (uid_t)VNOVAL) 588 uid = ip->i_uid; 589 if (gid == (gid_t)VNOVAL) 590 gid = ip->i_gid; 591 /* 592 * To modify the ownership of a file, must possess VADMIN 593 * for that file. 594 */ 595 if ((error = VOP_ACCESS(vp, VADMIN, cred, p))) 596 return (error); 597 /* 598 * To change the owner of a file, or change the group of a file 599 * to a group of which we are not a member, the caller must 600 * have privilege. 601 */ 602 if ((uid != ip->i_uid || 603 (gid != ip->i_gid && !groupmember(gid, cred))) && 604 (error = suser_xxx(cred, p, PRISON_ROOT))) 605 return (error); 606 ogid = ip->i_gid; 607 ouid = ip->i_uid; 608#ifdef QUOTA 609 if ((error = getinoquota(ip)) != 0) 610 return (error); 611 if (ouid == uid) { 612 dqrele(vp, ip->i_dquot[USRQUOTA]); 613 ip->i_dquot[USRQUOTA] = NODQUOT; 614 } 615 if (ogid == gid) { 616 dqrele(vp, ip->i_dquot[GRPQUOTA]); 617 ip->i_dquot[GRPQUOTA] = NODQUOT; 618 } 619 change = ip->i_blocks; 620 (void) chkdq(ip, -change, cred, CHOWN); 621 (void) chkiq(ip, -1, cred, CHOWN); 622 for (i = 0; i < MAXQUOTAS; i++) { 623 dqrele(vp, ip->i_dquot[i]); 624 ip->i_dquot[i] = NODQUOT; 625 } 626#endif 627 ip->i_gid = gid; 628 ip->i_uid = uid; 629#ifdef QUOTA 630 if ((error = getinoquota(ip)) == 0) { 631 if (ouid == uid) { 632 dqrele(vp, ip->i_dquot[USRQUOTA]); 633 ip->i_dquot[USRQUOTA] = NODQUOT; 634 } 635 if (ogid == gid) { 636 dqrele(vp, ip->i_dquot[GRPQUOTA]); 637 ip->i_dquot[GRPQUOTA] = NODQUOT; 638 } 639 if ((error = chkdq(ip, change, cred, CHOWN)) == 0) { 640 if ((error = chkiq(ip, 1, cred, CHOWN)) == 0) 641 goto good; 642 else 643 (void) chkdq(ip, -change, cred, CHOWN|FORCE); 644 } 645 for (i = 0; i < MAXQUOTAS; i++) { 646 dqrele(vp, ip->i_dquot[i]); 647 ip->i_dquot[i] = NODQUOT; 648 } 649 } 650 ip->i_gid = ogid; 651 ip->i_uid = ouid; 652 if (getinoquota(ip) == 0) { 653 if (ouid == uid) { 654 dqrele(vp, ip->i_dquot[USRQUOTA]); 655 ip->i_dquot[USRQUOTA] = NODQUOT; 656 } 657 if (ogid == gid) { 658 dqrele(vp, ip->i_dquot[GRPQUOTA]); 659 ip->i_dquot[GRPQUOTA] = NODQUOT; 660 } 661 (void) chkdq(ip, change, cred, FORCE|CHOWN); 662 (void) chkiq(ip, 1, cred, FORCE|CHOWN); 663 (void) getinoquota(ip); 664 } 665 return (error); 666good: 667 if (getinoquota(ip)) 668 panic("ufs_chown: lost quota"); 669#endif /* QUOTA */ 670 ip->i_flag |= IN_CHANGE; 671 if (suser_xxx(cred, NULL, PRISON_ROOT) && (ouid != uid || ogid != gid)) 672 ip->i_mode &= ~(ISUID | ISGID); 673 return (0); 674} 675 676/* 677 * Mmap a file 678 * 679 * NB Currently unsupported. 680 */ 681/* ARGSUSED */ 682int 683ufs_mmap(ap) 684 struct vop_mmap_args /* { 685 struct vnode *a_vp; 686 int a_fflags; 687 struct ucred *a_cred; 688 struct proc *a_p; 689 } */ *ap; 690{ 691 692 return (EINVAL); 693} 694 695int 696ufs_remove(ap) 697 struct vop_remove_args /* { 698 struct vnode *a_dvp; 699 struct vnode *a_vp; 700 struct componentname *a_cnp; 701 } */ *ap; 702{ 703 struct inode *ip; 704 struct vnode *vp = ap->a_vp; 705 struct vnode *dvp = ap->a_dvp; 706 int error; 707 708 ip = VTOI(vp); 709 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) || 710 (VTOI(dvp)->i_flags & APPEND)) { 711 error = EPERM; 712 goto out; 713 } 714 error = ufs_dirremove(dvp, ip, ap->a_cnp->cn_flags, 0); 715 VN_KNOTE(vp, NOTE_DELETE); 716 VN_KNOTE(dvp, NOTE_WRITE); 717out: 718 return (error); 719} 720 721/* 722 * link vnode call 723 */ 724int 725ufs_link(ap) 726 struct vop_link_args /* { 727 struct vnode *a_tdvp; 728 struct vnode *a_vp; 729 struct componentname *a_cnp; 730 } */ *ap; 731{ 732 struct vnode *vp = ap->a_vp; 733 struct vnode *tdvp = ap->a_tdvp; 734 struct componentname *cnp = ap->a_cnp; 735 struct proc *p = cnp->cn_proc; 736 struct inode *ip; 737 struct direct newdir; 738 int error; 739 740#ifdef DIAGNOSTIC 741 if ((cnp->cn_flags & HASBUF) == 0) 742 panic("ufs_link: no name"); 743#endif 744 if (tdvp->v_mount != vp->v_mount) { 745 error = EXDEV; 746 goto out2; 747 } 748 if (tdvp != vp && (error = vn_lock(vp, LK_EXCLUSIVE, p))) { 749 goto out2; 750 } 751 ip = VTOI(vp); 752 if ((nlink_t)ip->i_nlink >= LINK_MAX) { 753 error = EMLINK; 754 goto out1; 755 } 756 if (ip->i_flags & (IMMUTABLE | APPEND)) { 757 error = EPERM; 758 goto out1; 759 } 760 ip->i_effnlink++; 761 ip->i_nlink++; 762 ip->i_flag |= IN_CHANGE; 763 if (DOINGSOFTDEP(vp)) 764 softdep_change_linkcnt(ip); 765 error = UFS_UPDATE(vp, !(DOINGSOFTDEP(vp) | DOINGASYNC(vp))); 766 if (!error) { 767 ufs_makedirentry(ip, cnp, &newdir); 768 error = ufs_direnter(tdvp, vp, &newdir, cnp, NULL); 769 } 770 771 if (error) { 772 ip->i_effnlink--; 773 ip->i_nlink--; 774 ip->i_flag |= IN_CHANGE; 775 if (DOINGSOFTDEP(vp)) 776 softdep_change_linkcnt(ip); 777 } 778out1: 779 if (tdvp != vp) 780 VOP_UNLOCK(vp, 0, p); 781out2: 782 VN_KNOTE(vp, NOTE_LINK); 783 VN_KNOTE(tdvp, NOTE_WRITE); 784 return (error); 785} 786 787/* 788 * whiteout vnode call 789 */ 790int 791ufs_whiteout(ap) 792 struct vop_whiteout_args /* { 793 struct vnode *a_dvp; 794 struct componentname *a_cnp; 795 int a_flags; 796 } */ *ap; 797{ 798 struct vnode *dvp = ap->a_dvp; 799 struct componentname *cnp = ap->a_cnp; 800 struct direct newdir; 801 int error = 0; 802 803 switch (ap->a_flags) { 804 case LOOKUP: 805 /* 4.4 format directories support whiteout operations */ 806 if (dvp->v_mount->mnt_maxsymlinklen > 0) 807 return (0); 808 return (EOPNOTSUPP); 809 810 case CREATE: 811 /* create a new directory whiteout */ 812#ifdef DIAGNOSTIC 813 if ((cnp->cn_flags & SAVENAME) == 0) 814 panic("ufs_whiteout: missing name"); 815 if (dvp->v_mount->mnt_maxsymlinklen <= 0) 816 panic("ufs_whiteout: old format filesystem"); 817#endif 818 819 newdir.d_ino = WINO; 820 newdir.d_namlen = cnp->cn_namelen; 821 bcopy(cnp->cn_nameptr, newdir.d_name, (unsigned)cnp->cn_namelen + 1); 822 newdir.d_type = DT_WHT; 823 error = ufs_direnter(dvp, NULL, &newdir, cnp, NULL); 824 break; 825 826 case DELETE: 827 /* remove an existing directory whiteout */ 828#ifdef DIAGNOSTIC 829 if (dvp->v_mount->mnt_maxsymlinklen <= 0) 830 panic("ufs_whiteout: old format filesystem"); 831#endif 832 833 cnp->cn_flags &= ~DOWHITEOUT; 834 error = ufs_dirremove(dvp, NULL, cnp->cn_flags, 0); 835 break; 836 default: 837 panic("ufs_whiteout: unknown op"); 838 } 839 return (error); 840} 841 842/* 843 * Rename system call. 844 * rename("foo", "bar"); 845 * is essentially 846 * unlink("bar"); 847 * link("foo", "bar"); 848 * unlink("foo"); 849 * but ``atomically''. Can't do full commit without saving state in the 850 * inode on disk which isn't feasible at this time. Best we can do is 851 * always guarantee the target exists. 852 * 853 * Basic algorithm is: 854 * 855 * 1) Bump link count on source while we're linking it to the 856 * target. This also ensure the inode won't be deleted out 857 * from underneath us while we work (it may be truncated by 858 * a concurrent `trunc' or `open' for creation). 859 * 2) Link source to destination. If destination already exists, 860 * delete it first. 861 * 3) Unlink source reference to inode if still around. If a 862 * directory was moved and the parent of the destination 863 * is different from the source, patch the ".." entry in the 864 * directory. 865 */ 866int 867ufs_rename(ap) 868 struct vop_rename_args /* { 869 struct vnode *a_fdvp; 870 struct vnode *a_fvp; 871 struct componentname *a_fcnp; 872 struct vnode *a_tdvp; 873 struct vnode *a_tvp; 874 struct componentname *a_tcnp; 875 } */ *ap; 876{ 877 struct vnode *tvp = ap->a_tvp; 878 register struct vnode *tdvp = ap->a_tdvp; 879 struct vnode *fvp = ap->a_fvp; 880 struct vnode *fdvp = ap->a_fdvp; 881 struct componentname *tcnp = ap->a_tcnp; 882 struct componentname *fcnp = ap->a_fcnp; 883 struct proc *p = fcnp->cn_proc; 884 struct inode *ip, *xp, *dp; 885 struct direct newdir; 886 int doingdirectory = 0, oldparent = 0, newparent = 0; 887 int error = 0, ioflag; 888 889#ifdef DIAGNOSTIC 890 if ((tcnp->cn_flags & HASBUF) == 0 || 891 (fcnp->cn_flags & HASBUF) == 0) 892 panic("ufs_rename: no name"); 893#endif 894 /* 895 * Check for cross-device rename. 896 */ 897 if ((fvp->v_mount != tdvp->v_mount) || 898 (tvp && (fvp->v_mount != tvp->v_mount))) { 899 error = EXDEV; 900abortit: 901 if (tdvp == tvp) 902 vrele(tdvp); 903 else 904 vput(tdvp); 905 if (tvp) 906 vput(tvp); 907 vrele(fdvp); 908 vrele(fvp); 909 return (error); 910 } 911 912 if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) || 913 (VTOI(tdvp)->i_flags & APPEND))) { 914 error = EPERM; 915 goto abortit; 916 } 917 918 /* 919 * Check if just deleting a link name or if we've lost a race. 920 * If another process completes the same rename after we've looked 921 * up the source and have blocked looking up the target, then the 922 * source and target inodes may be identical now although the 923 * names were never linked. 924 */ 925 if (fvp == tvp) { 926 if (fvp->v_type == VDIR) { 927 /* 928 * Linked directories are impossible, so we must 929 * have lost the race. Pretend that the rename 930 * completed before the lookup. 931 */ 932#ifdef UFS_RENAME_DEBUG 933 printf("ufs_rename: fvp == tvp for directories\n"); 934#endif 935 error = ENOENT; 936 goto abortit; 937 } 938 939 /* Release destination completely. */ 940 vput(tdvp); 941 vput(tvp); 942 943 /* 944 * Delete source. There is another race now that everything 945 * is unlocked, but this doesn't cause any new complications. 946 * Relookup() may find a file that is unrelated to the 947 * original one, or it may fail. Too bad. 948 */ 949 vrele(fdvp); 950 vrele(fvp); 951 fcnp->cn_flags &= ~MODMASK; 952 fcnp->cn_flags |= LOCKPARENT | LOCKLEAF; 953 if ((fcnp->cn_flags & SAVESTART) == 0) 954 panic("ufs_rename: lost from startdir"); 955 fcnp->cn_nameiop = DELETE; 956 VREF(fdvp); 957 error = relookup(fdvp, &fvp, fcnp); 958 if (error == 0) 959 vrele(fdvp); 960 if (fvp == NULL) { 961#ifdef UFS_RENAME_DEBUG 962 printf("ufs_rename: from name disappeared\n"); 963#endif 964 return (ENOENT); 965 } 966 error = VOP_REMOVE(fdvp, fvp, fcnp); 967 if (fdvp == fvp) 968 vrele(fdvp); 969 else 970 vput(fdvp); 971 vput(fvp); 972 return (error); 973 } 974 if ((error = vn_lock(fvp, LK_EXCLUSIVE, p)) != 0) 975 goto abortit; 976 dp = VTOI(fdvp); 977 ip = VTOI(fvp); 978 if (ip->i_nlink >= LINK_MAX) { 979 VOP_UNLOCK(fvp, 0, p); 980 error = EMLINK; 981 goto abortit; 982 } 983 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) 984 || (dp->i_flags & APPEND)) { 985 VOP_UNLOCK(fvp, 0, p); 986 error = EPERM; 987 goto abortit; 988 } 989 if ((ip->i_mode & IFMT) == IFDIR) { 990 /* 991 * Avoid ".", "..", and aliases of "." for obvious reasons. 992 */ 993 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') || 994 dp == ip || (fcnp->cn_flags | tcnp->cn_flags) & ISDOTDOT || 995 (ip->i_flag & IN_RENAME)) { 996 VOP_UNLOCK(fvp, 0, p); 997 error = EINVAL; 998 goto abortit; 999 } 1000 ip->i_flag |= IN_RENAME; 1001 oldparent = dp->i_number; 1002 doingdirectory = 1; 1003 } 1004 VN_KNOTE(fdvp, NOTE_WRITE); /* XXX right place? */ 1005 vrele(fdvp); 1006 1007 /* 1008 * When the target exists, both the directory 1009 * and target vnodes are returned locked. 1010 */ 1011 dp = VTOI(tdvp); 1012 xp = NULL; 1013 if (tvp) 1014 xp = VTOI(tvp); 1015 1016 /* 1017 * 1) Bump link count while we're moving stuff 1018 * around. If we crash somewhere before 1019 * completing our work, the link count 1020 * may be wrong, but correctable. 1021 */ 1022 ip->i_effnlink++; 1023 ip->i_nlink++; 1024 ip->i_flag |= IN_CHANGE; 1025 if (DOINGSOFTDEP(fvp)) 1026 softdep_change_linkcnt(ip); 1027 if ((error = UFS_UPDATE(fvp, !(DOINGSOFTDEP(fvp) | 1028 DOINGASYNC(fvp)))) != 0) { 1029 VOP_UNLOCK(fvp, 0, p); 1030 goto bad; 1031 } 1032 1033 /* 1034 * If ".." must be changed (ie the directory gets a new 1035 * parent) then the source directory must not be in the 1036 * directory heirarchy above the target, as this would 1037 * orphan everything below the source directory. Also 1038 * the user must have write permission in the source so 1039 * as to be able to change "..". We must repeat the call 1040 * to namei, as the parent directory is unlocked by the 1041 * call to checkpath(). 1042 */ 1043 error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred, tcnp->cn_proc); 1044 VOP_UNLOCK(fvp, 0, p); 1045 if (oldparent != dp->i_number) 1046 newparent = dp->i_number; 1047 if (doingdirectory && newparent) { 1048 if (error) /* write access check above */ 1049 goto bad; 1050 if (xp != NULL) 1051 vput(tvp); 1052 error = ufs_checkpath(ip, dp, tcnp->cn_cred); 1053 if (error) 1054 goto out; 1055 if ((tcnp->cn_flags & SAVESTART) == 0) 1056 panic("ufs_rename: lost to startdir"); 1057 VREF(tdvp); 1058 error = relookup(tdvp, &tvp, tcnp); 1059 if (error) 1060 goto out; 1061 vrele(tdvp); 1062 dp = VTOI(tdvp); 1063 xp = NULL; 1064 if (tvp) 1065 xp = VTOI(tvp); 1066 } 1067 /* 1068 * 2) If target doesn't exist, link the target 1069 * to the source and unlink the source. 1070 * Otherwise, rewrite the target directory 1071 * entry to reference the source inode and 1072 * expunge the original entry's existence. 1073 */ 1074 if (xp == NULL) { 1075 if (dp->i_dev != ip->i_dev) 1076 panic("ufs_rename: EXDEV"); 1077 /* 1078 * Account for ".." in new directory. 1079 * When source and destination have the same 1080 * parent we don't fool with the link count. 1081 */ 1082 if (doingdirectory && newparent) { 1083 if ((nlink_t)dp->i_nlink >= LINK_MAX) { 1084 error = EMLINK; 1085 goto bad; 1086 } 1087 dp->i_effnlink++; 1088 dp->i_nlink++; 1089 dp->i_flag |= IN_CHANGE; 1090 if (DOINGSOFTDEP(tdvp)) 1091 softdep_change_linkcnt(dp); 1092 error = UFS_UPDATE(tdvp, !(DOINGSOFTDEP(tdvp) | 1093 DOINGASYNC(tdvp))); 1094 if (error) 1095 goto bad; 1096 } 1097 ufs_makedirentry(ip, tcnp, &newdir); 1098 error = ufs_direnter(tdvp, NULL, &newdir, tcnp, NULL); 1099 if (error) { 1100 if (doingdirectory && newparent) { 1101 dp->i_effnlink--; 1102 dp->i_nlink--; 1103 dp->i_flag |= IN_CHANGE; 1104 if (DOINGSOFTDEP(tdvp)) 1105 softdep_change_linkcnt(dp); 1106 (void)UFS_UPDATE(tdvp, 1); 1107 } 1108 goto bad; 1109 } 1110 VN_KNOTE(tdvp, NOTE_WRITE); 1111 vput(tdvp); 1112 } else { 1113 if (xp->i_dev != dp->i_dev || xp->i_dev != ip->i_dev) 1114 panic("ufs_rename: EXDEV"); 1115 /* 1116 * Short circuit rename(foo, foo). 1117 */ 1118 if (xp->i_number == ip->i_number) 1119 panic("ufs_rename: same file"); 1120 /* 1121 * If the parent directory is "sticky", then the caller 1122 * must possess VADMIN for the parent directory, or the 1123 * destination of the rename. This implements append-only 1124 * directories. 1125 */ 1126 if ((dp->i_mode & S_ISTXT) && 1127 VOP_ACCESS(tdvp, VADMIN, tcnp->cn_cred, p) && 1128 VOP_ACCESS(tvp, VADMIN, tcnp->cn_cred, p)) { 1129 error = EPERM; 1130 goto bad; 1131 } 1132 /* 1133 * Target must be empty if a directory and have no links 1134 * to it. Also, ensure source and target are compatible 1135 * (both directories, or both not directories). 1136 */ 1137 if ((xp->i_mode&IFMT) == IFDIR) { 1138 if ((xp->i_effnlink > 2) || 1139 !ufs_dirempty(xp, dp->i_number, tcnp->cn_cred)) { 1140 error = ENOTEMPTY; 1141 goto bad; 1142 } 1143 if (!doingdirectory) { 1144 error = ENOTDIR; 1145 goto bad; 1146 } 1147 cache_purge(tdvp); 1148 } else if (doingdirectory) { 1149 error = EISDIR; 1150 goto bad; 1151 } 1152 error = ufs_dirrewrite(dp, xp, ip->i_number, 1153 IFTODT(ip->i_mode), 1154 (doingdirectory && newparent) ? newparent : doingdirectory); 1155 if (error) 1156 goto bad; 1157 if (doingdirectory) { 1158 if (!newparent) { 1159 dp->i_effnlink--; 1160 if (DOINGSOFTDEP(tdvp)) 1161 softdep_change_linkcnt(dp); 1162 } 1163 xp->i_effnlink--; 1164 if (DOINGSOFTDEP(tvp)) 1165 softdep_change_linkcnt(xp); 1166 } 1167 if (doingdirectory && !DOINGSOFTDEP(tvp)) { 1168 /* 1169 * Truncate inode. The only stuff left in the directory 1170 * is "." and "..". The "." reference is inconsequential 1171 * since we are quashing it. We have removed the "." 1172 * reference and the reference in the parent directory, 1173 * but there may be other hard links. The soft 1174 * dependency code will arrange to do these operations 1175 * after the parent directory entry has been deleted on 1176 * disk, so when running with that code we avoid doing 1177 * them now. 1178 */ 1179 if (!newparent) { 1180 dp->i_nlink--; 1181 dp->i_flag |= IN_CHANGE; 1182 } 1183 xp->i_nlink--; 1184 xp->i_flag |= IN_CHANGE; 1185 ioflag = DOINGASYNC(tvp) ? 0 : IO_SYNC; 1186 if ((error = UFS_TRUNCATE(tvp, (off_t)0, ioflag, 1187 tcnp->cn_cred, tcnp->cn_proc)) != 0) 1188 goto bad; 1189 } 1190 VN_KNOTE(tdvp, NOTE_WRITE); 1191 vput(tdvp); 1192 VN_KNOTE(tvp, NOTE_DELETE); 1193 vput(tvp); 1194 xp = NULL; 1195 } 1196 1197 /* 1198 * 3) Unlink the source. 1199 */ 1200 fcnp->cn_flags &= ~MODMASK; 1201 fcnp->cn_flags |= LOCKPARENT | LOCKLEAF; 1202 if ((fcnp->cn_flags & SAVESTART) == 0) 1203 panic("ufs_rename: lost from startdir"); 1204 VREF(fdvp); 1205 error = relookup(fdvp, &fvp, fcnp); 1206 if (error == 0) 1207 vrele(fdvp); 1208 if (fvp != NULL) { 1209 xp = VTOI(fvp); 1210 dp = VTOI(fdvp); 1211 } else { 1212 /* 1213 * From name has disappeared. 1214 */ 1215 if (doingdirectory) 1216 panic("ufs_rename: lost dir entry"); 1217 vrele(ap->a_fvp); 1218 return (0); 1219 } 1220 /* 1221 * Ensure that the directory entry still exists and has not 1222 * changed while the new name has been entered. If the source is 1223 * a file then the entry may have been unlinked or renamed. In 1224 * either case there is no further work to be done. If the source 1225 * is a directory then it cannot have been rmdir'ed; the IN_RENAME 1226 * flag ensures that it cannot be moved by another rename or removed 1227 * by a rmdir. 1228 */ 1229 if (xp != ip) { 1230 if (doingdirectory) 1231 panic("ufs_rename: lost dir entry"); 1232 } else { 1233 /* 1234 * If the source is a directory with a 1235 * new parent, the link count of the old 1236 * parent directory must be decremented 1237 * and ".." set to point to the new parent. 1238 */ 1239 if (doingdirectory && newparent) { 1240 xp->i_offset = mastertemplate.dot_reclen; 1241 ufs_dirrewrite(xp, dp, newparent, DT_DIR, 0); 1242 cache_purge(fdvp); 1243 } 1244 error = ufs_dirremove(fdvp, xp, fcnp->cn_flags, 0); 1245 xp->i_flag &= ~IN_RENAME; 1246 } 1247 VN_KNOTE(fvp, NOTE_RENAME); 1248 if (dp) 1249 vput(fdvp); 1250 if (xp) 1251 vput(fvp); 1252 vrele(ap->a_fvp); 1253 return (error); 1254 1255bad: 1256 if (xp) 1257 vput(ITOV(xp)); 1258 vput(ITOV(dp)); 1259out: 1260 if (doingdirectory) 1261 ip->i_flag &= ~IN_RENAME; 1262 if (vn_lock(fvp, LK_EXCLUSIVE, p) == 0) { 1263 ip->i_effnlink--; 1264 ip->i_nlink--; 1265 ip->i_flag |= IN_CHANGE; 1266 ip->i_flag &= ~IN_RENAME; 1267 if (DOINGSOFTDEP(fvp)) 1268 softdep_change_linkcnt(ip); 1269 vput(fvp); 1270 } else 1271 vrele(fvp); 1272 return (error); 1273} 1274 1275/* 1276 * Mkdir system call 1277 */ 1278int 1279ufs_mkdir(ap) 1280 struct vop_mkdir_args /* { 1281 struct vnode *a_dvp; 1282 struct vnode **a_vpp; 1283 struct componentname *a_cnp; 1284 struct vattr *a_vap; 1285 } */ *ap; 1286{ 1287 register struct vnode *dvp = ap->a_dvp; 1288 register struct vattr *vap = ap->a_vap; 1289 register struct componentname *cnp = ap->a_cnp; 1290 register struct inode *ip, *dp; 1291 struct vnode *tvp; 1292 struct buf *bp; 1293 struct dirtemplate dirtemplate, *dtp; 1294 struct direct newdir; 1295 int error, dmode; 1296 long blkoff; 1297 1298#ifdef DIAGNOSTIC 1299 if ((cnp->cn_flags & HASBUF) == 0) 1300 panic("ufs_mkdir: no name"); 1301#endif 1302 dp = VTOI(dvp); 1303 if ((nlink_t)dp->i_nlink >= LINK_MAX) { 1304 error = EMLINK; 1305 goto out; 1306 } 1307 dmode = vap->va_mode & 0777; 1308 dmode |= IFDIR; 1309 /* 1310 * Must simulate part of ufs_makeinode here to acquire the inode, 1311 * but not have it entered in the parent directory. The entry is 1312 * made later after writing "." and ".." entries. 1313 */ 1314 error = UFS_VALLOC(dvp, dmode, cnp->cn_cred, &tvp); 1315 if (error) 1316 goto out; 1317 ip = VTOI(tvp); 1318 ip->i_gid = dp->i_gid; 1319#ifdef SUIDDIR 1320 { 1321#ifdef QUOTA 1322 struct ucred ucred, *ucp; 1323 ucp = cnp->cn_cred; 1324#endif 1325 /* 1326 * If we are hacking owners here, (only do this where told to) 1327 * and we are not giving it TO root, (would subvert quotas) 1328 * then go ahead and give it to the other user. 1329 * The new directory also inherits the SUID bit. 1330 * If user's UID and dir UID are the same, 1331 * 'give it away' so that the SUID is still forced on. 1332 */ 1333 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) && 1334 (dp->i_mode & ISUID) && dp->i_uid) { 1335 dmode |= ISUID; 1336 ip->i_uid = dp->i_uid; 1337#ifdef QUOTA 1338 if (dp->i_uid != cnp->cn_cred->cr_uid) { 1339 /* 1340 * Make sure the correct user gets charged 1341 * for the space. 1342 * Make a dummy credential for the victim. 1343 * XXX This seems to never be accessed out of 1344 * our context so a stack variable is ok. 1345 */ 1346 ucred.cr_ref = 1; 1347 ucred.cr_uid = ip->i_uid; 1348 ucred.cr_ngroups = 1; 1349 ucred.cr_groups[0] = dp->i_gid; 1350 ucp = &ucred; 1351 } 1352#endif 1353 } else 1354 ip->i_uid = cnp->cn_cred->cr_uid; 1355#ifdef QUOTA 1356 if ((error = getinoquota(ip)) || 1357 (error = chkiq(ip, 1, ucp, 0))) { 1358 UFS_VFREE(tvp, ip->i_number, dmode); 1359 vput(tvp); 1360 return (error); 1361 } 1362#endif 1363 } 1364#else /* !SUIDDIR */ 1365 ip->i_uid = cnp->cn_cred->cr_uid; 1366#ifdef QUOTA 1367 if ((error = getinoquota(ip)) || 1368 (error = chkiq(ip, 1, cnp->cn_cred, 0))) { 1369 UFS_VFREE(tvp, ip->i_number, dmode); 1370 vput(tvp); 1371 return (error); 1372 } 1373#endif 1374#endif /* !SUIDDIR */ 1375 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE; 1376 ip->i_mode = dmode; 1377 tvp->v_type = VDIR; /* Rest init'd in getnewvnode(). */ 1378 ip->i_effnlink = 2; 1379 ip->i_nlink = 2; 1380 if (DOINGSOFTDEP(tvp)) 1381 softdep_change_linkcnt(ip); 1382 if (cnp->cn_flags & ISWHITEOUT) 1383 ip->i_flags |= UF_OPAQUE; 1384 1385 /* 1386 * Bump link count in parent directory to reflect work done below. 1387 * Should be done before reference is created so cleanup is 1388 * possible if we crash. 1389 */ 1390 dp->i_effnlink++; 1391 dp->i_nlink++; 1392 dp->i_flag |= IN_CHANGE; 1393 if (DOINGSOFTDEP(dvp)) 1394 softdep_change_linkcnt(dp); 1395 error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(dvp) | DOINGASYNC(dvp))); 1396 if (error) 1397 goto bad; 1398 1399 /* 1400 * Initialize directory with "." and ".." from static template. 1401 */ 1402 if (dvp->v_mount->mnt_maxsymlinklen > 0 1403 ) 1404 dtp = &mastertemplate; 1405 else 1406 dtp = (struct dirtemplate *)&omastertemplate; 1407 dirtemplate = *dtp; 1408 dirtemplate.dot_ino = ip->i_number; 1409 dirtemplate.dotdot_ino = dp->i_number; 1410 if ((error = VOP_BALLOC(tvp, (off_t)0, DIRBLKSIZ, cnp->cn_cred, 1411 B_CLRBUF, &bp)) != 0) 1412 goto bad; 1413 ip->i_size = DIRBLKSIZ; 1414 ip->i_flag |= IN_CHANGE | IN_UPDATE; 1415 vnode_pager_setsize(tvp, (u_long)ip->i_size); 1416 bcopy((caddr_t)&dirtemplate, (caddr_t)bp->b_data, sizeof dirtemplate); 1417 if (DOINGSOFTDEP(tvp)) { 1418 /* 1419 * Ensure that the entire newly allocated block is a 1420 * valid directory so that future growth within the 1421 * block does not have to ensure that the block is 1422 * written before the inode. 1423 */ 1424 blkoff = DIRBLKSIZ; 1425 while (blkoff < bp->b_bcount) { 1426 ((struct direct *) 1427 (bp->b_data + blkoff))->d_reclen = DIRBLKSIZ; 1428 blkoff += DIRBLKSIZ; 1429 } 1430 } 1431 if ((error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) | 1432 DOINGASYNC(tvp)))) != 0) { 1433 (void)BUF_WRITE(bp); 1434 goto bad; 1435 } 1436 /* 1437 * Directory set up, now install its entry in the parent directory. 1438 * 1439 * If we are not doing soft dependencies, then we must write out the 1440 * buffer containing the new directory body before entering the new 1441 * name in the parent. If we are doing soft dependencies, then the 1442 * buffer containing the new directory body will be passed to and 1443 * released in the soft dependency code after the code has attached 1444 * an appropriate ordering dependency to the buffer which ensures that 1445 * the buffer is written before the new name is written in the parent. 1446 */ 1447 if (DOINGASYNC(dvp)) 1448 bdwrite(bp); 1449 else if (!DOINGSOFTDEP(dvp) && ((error = BUF_WRITE(bp)))) 1450 goto bad; 1451 ufs_makedirentry(ip, cnp, &newdir); 1452 error = ufs_direnter(dvp, tvp, &newdir, cnp, bp); 1453 1454bad: 1455 if (error == 0) { 1456 VN_KNOTE(dvp, NOTE_WRITE); 1457 *ap->a_vpp = tvp; 1458 } else { 1459 dp->i_effnlink--; 1460 dp->i_nlink--; 1461 dp->i_flag |= IN_CHANGE; 1462 if (DOINGSOFTDEP(dvp)) 1463 softdep_change_linkcnt(dp); 1464 /* 1465 * No need to do an explicit VOP_TRUNCATE here, vrele will 1466 * do this for us because we set the link count to 0. 1467 */ 1468 ip->i_effnlink = 0; 1469 ip->i_nlink = 0; 1470 ip->i_flag |= IN_CHANGE; 1471 if (DOINGSOFTDEP(tvp)) 1472 softdep_change_linkcnt(ip); 1473 vput(tvp); 1474 } 1475out: 1476 return (error); 1477} 1478 1479/* 1480 * Rmdir system call. 1481 */ 1482int 1483ufs_rmdir(ap) 1484 struct vop_rmdir_args /* { 1485 struct vnode *a_dvp; 1486 struct vnode *a_vp; 1487 struct componentname *a_cnp; 1488 } */ *ap; 1489{ 1490 struct vnode *vp = ap->a_vp; 1491 struct vnode *dvp = ap->a_dvp; 1492 struct componentname *cnp = ap->a_cnp; 1493 struct inode *ip, *dp; 1494 int error, ioflag; 1495 1496 ip = VTOI(vp); 1497 dp = VTOI(dvp); 1498 1499 /* 1500 * Do not remove a directory that is in the process of being renamed. 1501 * Verify the directory is empty (and valid). Rmdir ".." will not be 1502 * valid since ".." will contain a reference to the current directory 1503 * and thus be non-empty. Do not allow the removal of mounted on 1504 * directories (this can happen when an NFS exported filesystem 1505 * tries to remove a locally mounted on directory). 1506 */ 1507 error = 0; 1508 if (ip->i_flag & IN_RENAME) { 1509 error = EINVAL; 1510 goto out; 1511 } 1512 if (ip->i_effnlink != 2 || 1513 !ufs_dirempty(ip, dp->i_number, cnp->cn_cred)) { 1514 error = ENOTEMPTY; 1515 goto out; 1516 } 1517 if ((dp->i_flags & APPEND) 1518 || (ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))) { 1519 error = EPERM; 1520 goto out; 1521 } 1522 if (vp->v_mountedhere != 0) { 1523 error = EINVAL; 1524 goto out; 1525 } 1526 /* 1527 * Delete reference to directory before purging 1528 * inode. If we crash in between, the directory 1529 * will be reattached to lost+found, 1530 */ 1531 dp->i_effnlink--; 1532 ip->i_effnlink--; 1533 if (DOINGSOFTDEP(vp)) { 1534 softdep_change_linkcnt(dp); 1535 softdep_change_linkcnt(ip); 1536 } 1537 error = ufs_dirremove(dvp, ip, cnp->cn_flags, 1); 1538 if (error) { 1539 dp->i_effnlink++; 1540 ip->i_effnlink++; 1541 if (DOINGSOFTDEP(vp)) { 1542 softdep_change_linkcnt(dp); 1543 softdep_change_linkcnt(ip); 1544 } 1545 goto out; 1546 } 1547 VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK); 1548 cache_purge(dvp); 1549 /* 1550 * Truncate inode. The only stuff left in the directory is "." and 1551 * "..". The "." reference is inconsequential since we are quashing 1552 * it. The soft dependency code will arrange to do these operations 1553 * after the parent directory entry has been deleted on disk, so 1554 * when running with that code we avoid doing them now. 1555 */ 1556 if (!DOINGSOFTDEP(vp)) { 1557 dp->i_nlink--; 1558 dp->i_flag |= IN_CHANGE; 1559 ip->i_nlink--; 1560 ip->i_flag |= IN_CHANGE; 1561 ioflag = DOINGASYNC(vp) ? 0 : IO_SYNC; 1562 error = UFS_TRUNCATE(vp, (off_t)0, ioflag, cnp->cn_cred, 1563 cnp->cn_proc); 1564 } 1565 cache_purge(vp); 1566out: 1567 VN_KNOTE(vp, NOTE_DELETE); 1568 return (error); 1569} 1570 1571/* 1572 * symlink -- make a symbolic link 1573 */ 1574int 1575ufs_symlink(ap) 1576 struct vop_symlink_args /* { 1577 struct vnode *a_dvp; 1578 struct vnode **a_vpp; 1579 struct componentname *a_cnp; 1580 struct vattr *a_vap; 1581 char *a_target; 1582 } */ *ap; 1583{ 1584 register struct vnode *vp, **vpp = ap->a_vpp; 1585 register struct inode *ip; 1586 int len, error; 1587 1588 error = ufs_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp, 1589 vpp, ap->a_cnp); 1590 if (error) 1591 return (error); 1592 VN_KNOTE(ap->a_dvp, NOTE_WRITE); 1593 vp = *vpp; 1594 len = strlen(ap->a_target); 1595 if (len < vp->v_mount->mnt_maxsymlinklen) { 1596 ip = VTOI(vp); 1597 bcopy(ap->a_target, (char *)ip->i_shortlink, len); 1598 ip->i_size = len; 1599 ip->i_flag |= IN_CHANGE | IN_UPDATE; 1600 } else 1601 error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0, 1602 UIO_SYSSPACE, IO_NODELOCKED, ap->a_cnp->cn_cred, (int *)0, 1603 (struct proc *)0); 1604 if (error) 1605 vput(vp); 1606 return (error); 1607} 1608 1609/* 1610 * Vnode op for reading directories. 1611 * 1612 * The routine below assumes that the on-disk format of a directory 1613 * is the same as that defined by <sys/dirent.h>. If the on-disk 1614 * format changes, then it will be necessary to do a conversion 1615 * from the on-disk format that read returns to the format defined 1616 * by <sys/dirent.h>. 1617 */ 1618int 1619ufs_readdir(ap) 1620 struct vop_readdir_args /* { 1621 struct vnode *a_vp; 1622 struct uio *a_uio; 1623 struct ucred *a_cred; 1624 int *a_eofflag; 1625 int *ncookies; 1626 u_long **a_cookies; 1627 } */ *ap; 1628{ 1629 register struct uio *uio = ap->a_uio; 1630 int error; 1631 size_t count, lost; 1632 off_t off; 1633 1634 if (ap->a_ncookies != NULL) 1635 /* 1636 * Ensure that the block is aligned. The caller can use 1637 * the cookies to determine where in the block to start. 1638 */ 1639 uio->uio_offset &= ~(DIRBLKSIZ - 1); 1640 off = uio->uio_offset; 1641 count = uio->uio_resid; 1642 /* Make sure we don't return partial entries. */ 1643 if (count <= ((uio->uio_offset + count) & (DIRBLKSIZ -1))) 1644 return (EINVAL); 1645 count -= (uio->uio_offset + count) & (DIRBLKSIZ -1); 1646 lost = uio->uio_resid - count; 1647 uio->uio_resid = count; 1648 uio->uio_iov->iov_len = count; 1649# if (BYTE_ORDER == LITTLE_ENDIAN) 1650 if (ap->a_vp->v_mount->mnt_maxsymlinklen > 0) { 1651 error = VOP_READ(ap->a_vp, uio, 0, ap->a_cred); 1652 } else { 1653 struct dirent *dp, *edp; 1654 struct uio auio; 1655 struct iovec aiov; 1656 caddr_t dirbuf; 1657 int readcnt; 1658 u_char tmp; 1659 1660 auio = *uio; 1661 auio.uio_iov = &aiov; 1662 auio.uio_iovcnt = 1; 1663 auio.uio_segflg = UIO_SYSSPACE; 1664 aiov.iov_len = count; 1665 MALLOC(dirbuf, caddr_t, count, M_TEMP, M_WAITOK); 1666 aiov.iov_base = dirbuf; 1667 error = VOP_READ(ap->a_vp, &auio, 0, ap->a_cred); 1668 if (error == 0) { 1669 readcnt = count - auio.uio_resid; 1670 edp = (struct dirent *)&dirbuf[readcnt]; 1671 for (dp = (struct dirent *)dirbuf; dp < edp; ) { 1672 tmp = dp->d_namlen; 1673 dp->d_namlen = dp->d_type; 1674 dp->d_type = tmp; 1675 if (dp->d_reclen > 0) { 1676 dp = (struct dirent *) 1677 ((char *)dp + dp->d_reclen); 1678 } else { 1679 error = EIO; 1680 break; 1681 } 1682 } 1683 if (dp >= edp) 1684 error = uiomove(dirbuf, readcnt, uio); 1685 } 1686 FREE(dirbuf, M_TEMP); 1687 } 1688# else 1689 error = VOP_READ(ap->a_vp, uio, 0, ap->a_cred); 1690# endif 1691 if (!error && ap->a_ncookies != NULL) { 1692 struct dirent* dpStart; 1693 struct dirent* dpEnd; 1694 struct dirent* dp; 1695 int ncookies; 1696 u_long *cookies; 1697 u_long *cookiep; 1698 1699 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) 1700 panic("ufs_readdir: unexpected uio from NFS server"); 1701 dpStart = (struct dirent *) 1702 (uio->uio_iov->iov_base - (uio->uio_offset - off)); 1703 dpEnd = (struct dirent *) uio->uio_iov->iov_base; 1704 for (dp = dpStart, ncookies = 0; 1705 dp < dpEnd; 1706 dp = (struct dirent *)((caddr_t) dp + dp->d_reclen)) 1707 ncookies++; 1708 MALLOC(cookies, u_long *, ncookies * sizeof(u_long), M_TEMP, 1709 M_WAITOK); 1710 for (dp = dpStart, cookiep = cookies; 1711 dp < dpEnd; 1712 dp = (struct dirent *)((caddr_t) dp + dp->d_reclen)) { 1713 off += dp->d_reclen; 1714 *cookiep++ = (u_long) off; 1715 } 1716 *ap->a_ncookies = ncookies; 1717 *ap->a_cookies = cookies; 1718 } 1719 uio->uio_resid += lost; 1720 if (ap->a_eofflag) 1721 *ap->a_eofflag = VTOI(ap->a_vp)->i_size <= uio->uio_offset; 1722 return (error); 1723} 1724 1725/* 1726 * Return target name of a symbolic link 1727 */ 1728int 1729ufs_readlink(ap) 1730 struct vop_readlink_args /* { 1731 struct vnode *a_vp; 1732 struct uio *a_uio; 1733 struct ucred *a_cred; 1734 } */ *ap; 1735{ 1736 register struct vnode *vp = ap->a_vp; 1737 register struct inode *ip = VTOI(vp); 1738 int isize; 1739 1740 isize = ip->i_size; 1741 if ((isize < vp->v_mount->mnt_maxsymlinklen) || 1742 (ip->i_din.di_blocks == 0)) { /* XXX - for old fastlink support */ 1743 uiomove((char *)ip->i_shortlink, isize, ap->a_uio); 1744 return (0); 1745 } 1746 return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred)); 1747} 1748 1749/* 1750 * Calculate the logical to physical mapping if not done already, 1751 * then call the device strategy routine. 1752 * 1753 * In order to be able to swap to a file, the VOP_BMAP operation may not 1754 * deadlock on memory. See ufs_bmap() for details. 1755 */ 1756int 1757ufs_strategy(ap) 1758 struct vop_strategy_args /* { 1759 struct vnode *a_vp; 1760 struct buf *a_bp; 1761 } */ *ap; 1762{ 1763 register struct buf *bp = ap->a_bp; 1764 register struct vnode *vp = ap->a_vp; 1765 register struct inode *ip; 1766 int error; 1767 1768 ip = VTOI(vp); 1769 if (vp->v_type == VBLK || vp->v_type == VCHR) 1770 panic("ufs_strategy: spec"); 1771 if (bp->b_blkno == bp->b_lblkno) { 1772 error = VOP_BMAP(vp, bp->b_lblkno, NULL, &bp->b_blkno, NULL, NULL); 1773 if (error) { 1774 bp->b_error = error; 1775 bp->b_ioflags |= BIO_ERROR; 1776 bufdone(bp); 1777 return (error); 1778 } 1779 if ((long)bp->b_blkno == -1) 1780 vfs_bio_clrbuf(bp); 1781 } 1782 if ((long)bp->b_blkno == -1) { 1783 bufdone(bp); 1784 return (0); 1785 } 1786 vp = ip->i_devvp; 1787 bp->b_dev = vp->v_rdev; 1788 VOP_STRATEGY(vp, bp); 1789 return (0); 1790} 1791 1792/* 1793 * Print out the contents of an inode. 1794 */ 1795int 1796ufs_print(ap) 1797 struct vop_print_args /* { 1798 struct vnode *a_vp; 1799 } */ *ap; 1800{ 1801 register struct vnode *vp = ap->a_vp; 1802 register struct inode *ip = VTOI(vp); 1803 1804 printf("tag VT_UFS, ino %lu, on dev %s (%d, %d)", 1805 (u_long)ip->i_number, devtoname(ip->i_dev), major(ip->i_dev), 1806 minor(ip->i_dev)); 1807 if (vp->v_type == VFIFO) 1808 fifo_printinfo(vp); 1809 lockmgr_printinfo(&vp->v_lock); 1810 printf("\n"); 1811 return (0); 1812} 1813 1814/* 1815 * Read wrapper for special devices. 1816 */ 1817int 1818ufsspec_read(ap) 1819 struct vop_read_args /* { 1820 struct vnode *a_vp; 1821 struct uio *a_uio; 1822 int a_ioflag; 1823 struct ucred *a_cred; 1824 } */ *ap; 1825{ 1826 int error, resid; 1827 struct inode *ip; 1828 struct uio *uio; 1829 1830 uio = ap->a_uio; 1831 resid = uio->uio_resid; 1832 error = VOCALL(spec_vnodeop_p, VOFFSET(vop_read), ap); 1833 /* 1834 * The inode may have been revoked during the call, so it must not 1835 * be accessed blindly here or in the other wrapper functions. 1836 */ 1837 ip = VTOI(ap->a_vp); 1838 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0))) 1839 ip->i_flag |= IN_ACCESS; 1840 return (error); 1841} 1842 1843/* 1844 * Write wrapper for special devices. 1845 */ 1846int 1847ufsspec_write(ap) 1848 struct vop_write_args /* { 1849 struct vnode *a_vp; 1850 struct uio *a_uio; 1851 int a_ioflag; 1852 struct ucred *a_cred; 1853 } */ *ap; 1854{ 1855 int error, resid; 1856 struct inode *ip; 1857 struct uio *uio; 1858 1859 uio = ap->a_uio; 1860 resid = uio->uio_resid; 1861 error = VOCALL(spec_vnodeop_p, VOFFSET(vop_write), ap); 1862 ip = VTOI(ap->a_vp); 1863 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0))) 1864 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE; 1865 return (error); 1866} 1867 1868/* 1869 * Close wrapper for special devices. 1870 * 1871 * Update the times on the inode then do device close. 1872 */ 1873int 1874ufsspec_close(ap) 1875 struct vop_close_args /* { 1876 struct vnode *a_vp; 1877 int a_fflag; 1878 struct ucred *a_cred; 1879 struct proc *a_p; 1880 } */ *ap; 1881{ 1882 struct vnode *vp = ap->a_vp; 1883 1884 mtx_enter(&vp->v_interlock, MTX_DEF); 1885 if (vp->v_usecount > 1) 1886 ufs_itimes(vp); 1887 mtx_exit(&vp->v_interlock, MTX_DEF); 1888 return (VOCALL(spec_vnodeop_p, VOFFSET(vop_close), ap)); 1889} 1890 1891/* 1892 * Read wrapper for fifos. 1893 */ 1894int 1895ufsfifo_read(ap) 1896 struct vop_read_args /* { 1897 struct vnode *a_vp; 1898 struct uio *a_uio; 1899 int a_ioflag; 1900 struct ucred *a_cred; 1901 } */ *ap; 1902{ 1903 int error, resid; 1904 struct inode *ip; 1905 struct uio *uio; 1906 1907 uio = ap->a_uio; 1908 resid = uio->uio_resid; 1909 error = VOCALL(fifo_vnodeop_p, VOFFSET(vop_read), ap); 1910 ip = VTOI(ap->a_vp); 1911 if ((ap->a_vp->v_mount->mnt_flag & MNT_NOATIME) == 0 && ip != NULL && 1912 (uio->uio_resid != resid || (error == 0 && resid != 0))) 1913 VTOI(ap->a_vp)->i_flag |= IN_ACCESS; 1914 return (error); 1915} 1916 1917/* 1918 * Write wrapper for fifos. 1919 */ 1920int 1921ufsfifo_write(ap) 1922 struct vop_write_args /* { 1923 struct vnode *a_vp; 1924 struct uio *a_uio; 1925 int a_ioflag; 1926 struct ucred *a_cred; 1927 } */ *ap; 1928{ 1929 int error, resid; 1930 struct inode *ip; 1931 struct uio *uio; 1932 1933 uio = ap->a_uio; 1934 resid = uio->uio_resid; 1935 error = VOCALL(fifo_vnodeop_p, VOFFSET(vop_write), ap); 1936 ip = VTOI(ap->a_vp); 1937 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0))) 1938 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE; 1939 return (error); 1940} 1941 1942/* 1943 * Close wrapper for fifos. 1944 * 1945 * Update the times on the inode then do device close. 1946 */ 1947int 1948ufsfifo_close(ap) 1949 struct vop_close_args /* { 1950 struct vnode *a_vp; 1951 int a_fflag; 1952 struct ucred *a_cred; 1953 struct proc *a_p; 1954 } */ *ap; 1955{ 1956 struct vnode *vp = ap->a_vp; 1957 1958 mtx_enter(&vp->v_interlock, MTX_DEF); 1959 if (vp->v_usecount > 1) 1960 ufs_itimes(vp); 1961 mtx_exit(&vp->v_interlock, MTX_DEF); 1962 return (VOCALL(fifo_vnodeop_p, VOFFSET(vop_close), ap)); 1963} 1964 1965/* 1966 * Return POSIX pathconf information applicable to ufs filesystems. 1967 */ 1968int 1969ufs_pathconf(ap) 1970 struct vop_pathconf_args /* { 1971 struct vnode *a_vp; 1972 int a_name; 1973 int *a_retval; 1974 } */ *ap; 1975{ 1976 1977 switch (ap->a_name) { 1978 case _PC_LINK_MAX: 1979 *ap->a_retval = LINK_MAX; 1980 return (0); 1981 case _PC_NAME_MAX: 1982 *ap->a_retval = NAME_MAX; 1983 return (0); 1984 case _PC_PATH_MAX: 1985 *ap->a_retval = PATH_MAX; 1986 return (0); 1987 case _PC_PIPE_BUF: 1988 *ap->a_retval = PIPE_BUF; 1989 return (0); 1990 case _PC_CHOWN_RESTRICTED: 1991 *ap->a_retval = 1; 1992 return (0); 1993 case _PC_NO_TRUNC: 1994 *ap->a_retval = 1; 1995 return (0); 1996 default: 1997 return (EINVAL); 1998 } 1999 /* NOTREACHED */ 2000} 2001 2002/* 2003 * Advisory record locking support 2004 */ 2005int 2006ufs_advlock(ap) 2007 struct vop_advlock_args /* { 2008 struct vnode *a_vp; 2009 caddr_t a_id; 2010 int a_op; 2011 struct flock *a_fl; 2012 int a_flags; 2013 } */ *ap; 2014{ 2015 register struct inode *ip = VTOI(ap->a_vp); 2016 2017 return (lf_advlock(ap, &(ip->i_lockf), ip->i_size)); 2018} 2019 2020/* 2021 * Initialize the vnode associated with a new inode, handle aliased 2022 * vnodes. 2023 */ 2024int 2025ufs_vinit(mntp, specops, fifoops, vpp) 2026 struct mount *mntp; 2027 vop_t **specops; 2028 vop_t **fifoops; 2029 struct vnode **vpp; 2030{ 2031 struct inode *ip; 2032 struct vnode *vp; 2033 struct timeval tv; 2034 2035 vp = *vpp; 2036 ip = VTOI(vp); 2037 switch(vp->v_type = IFTOVT(ip->i_mode)) { 2038 case VCHR: 2039 case VBLK: 2040 vp->v_op = specops; 2041 vp = addaliasu(vp, ip->i_rdev); 2042 ip->i_vnode = vp; 2043 break; 2044 case VFIFO: 2045 vp->v_op = fifoops; 2046 break; 2047 default: 2048 break; 2049 2050 } 2051 if (ip->i_number == ROOTINO) 2052 vp->v_flag |= VROOT; 2053 /* 2054 * Initialize modrev times 2055 */ 2056 getmicrouptime(&tv); 2057 SETHIGH(ip->i_modrev, tv.tv_sec); 2058 SETLOW(ip->i_modrev, tv.tv_usec * 4294); 2059 *vpp = vp; 2060 return (0); 2061} 2062 2063/* 2064 * Allocate a new inode. 2065 */ 2066int 2067ufs_makeinode(mode, dvp, vpp, cnp) 2068 int mode; 2069 struct vnode *dvp; 2070 struct vnode **vpp; 2071 struct componentname *cnp; 2072{ 2073 register struct inode *ip, *pdir; 2074 struct direct newdir; 2075 struct vnode *tvp; 2076 int error; 2077 2078 pdir = VTOI(dvp); 2079#ifdef DIAGNOSTIC 2080 if ((cnp->cn_flags & HASBUF) == 0) 2081 panic("ufs_makeinode: no name"); 2082#endif 2083 *vpp = NULL; 2084 if ((mode & IFMT) == 0) 2085 mode |= IFREG; 2086 2087 error = UFS_VALLOC(dvp, mode, cnp->cn_cred, &tvp); 2088 if (error) 2089 return (error); 2090 ip = VTOI(tvp); 2091 ip->i_gid = pdir->i_gid; 2092#ifdef SUIDDIR 2093 { 2094#ifdef QUOTA 2095 struct ucred ucred, *ucp; 2096 ucp = cnp->cn_cred; 2097#endif 2098 /* 2099 * If we are not the owner of the directory, 2100 * and we are hacking owners here, (only do this where told to) 2101 * and we are not giving it TO root, (would subvert quotas) 2102 * then go ahead and give it to the other user. 2103 * Note that this drops off the execute bits for security. 2104 */ 2105 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) && 2106 (pdir->i_mode & ISUID) && 2107 (pdir->i_uid != cnp->cn_cred->cr_uid) && pdir->i_uid) { 2108 ip->i_uid = pdir->i_uid; 2109 mode &= ~07111; 2110#ifdef QUOTA 2111 /* 2112 * Make sure the correct user gets charged 2113 * for the space. 2114 * Quickly knock up a dummy credential for the victim. 2115 * XXX This seems to never be accessed out of our 2116 * context so a stack variable is ok. 2117 */ 2118 ucred.cr_ref = 1; 2119 ucred.cr_uid = ip->i_uid; 2120 ucred.cr_ngroups = 1; 2121 ucred.cr_groups[0] = pdir->i_gid; 2122 ucp = &ucred; 2123#endif 2124 } else 2125 ip->i_uid = cnp->cn_cred->cr_uid; 2126 2127#ifdef QUOTA 2128 if ((error = getinoquota(ip)) || 2129 (error = chkiq(ip, 1, ucp, 0))) { 2130 UFS_VFREE(tvp, ip->i_number, mode); 2131 vput(tvp); 2132 return (error); 2133 } 2134#endif 2135 } 2136#else /* !SUIDDIR */ 2137 ip->i_uid = cnp->cn_cred->cr_uid; 2138#ifdef QUOTA 2139 if ((error = getinoquota(ip)) || 2140 (error = chkiq(ip, 1, cnp->cn_cred, 0))) { 2141 UFS_VFREE(tvp, ip->i_number, mode); 2142 vput(tvp); 2143 return (error); 2144 } 2145#endif 2146#endif /* !SUIDDIR */ 2147 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE; 2148 ip->i_mode = mode; 2149 tvp->v_type = IFTOVT(mode); /* Rest init'd in getnewvnode(). */ 2150 ip->i_effnlink = 1; 2151 ip->i_nlink = 1; 2152 if (DOINGSOFTDEP(tvp)) 2153 softdep_change_linkcnt(ip); 2154 if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, cnp->cn_cred) && 2155 suser_xxx(cnp->cn_cred, NULL, PRISON_ROOT)) 2156 ip->i_mode &= ~ISGID; 2157 2158 if (cnp->cn_flags & ISWHITEOUT) 2159 ip->i_flags |= UF_OPAQUE; 2160 2161 /* 2162 * Make sure inode goes to disk before directory entry. 2163 */ 2164 error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) | DOINGASYNC(tvp))); 2165 if (error) 2166 goto bad; 2167 ufs_makedirentry(ip, cnp, &newdir); 2168 error = ufs_direnter(dvp, tvp, &newdir, cnp, NULL); 2169 if (error) 2170 goto bad; 2171 *vpp = tvp; 2172 return (0); 2173 2174bad: 2175 /* 2176 * Write error occurred trying to update the inode 2177 * or the directory so must deallocate the inode. 2178 */ 2179 ip->i_effnlink = 0; 2180 ip->i_nlink = 0; 2181 ip->i_flag |= IN_CHANGE; 2182 if (DOINGSOFTDEP(tvp)) 2183 softdep_change_linkcnt(ip); 2184 vput(tvp); 2185 return (error); 2186} 2187 2188static int 2189ufs_missingop(ap) 2190 struct vop_generic_args *ap; 2191{ 2192 2193 panic("no vop function for %s in ufs child", ap->a_desc->vdesc_name); 2194 return (EOPNOTSUPP); 2195} 2196 2197/* Global vfs data structures for ufs. */ 2198static vop_t **ufs_vnodeop_p; 2199static struct vnodeopv_entry_desc ufs_vnodeop_entries[] = { 2200 { &vop_default_desc, (vop_t *) vop_defaultop }, 2201 { &vop_fsync_desc, (vop_t *) ufs_missingop }, 2202 { &vop_read_desc, (vop_t *) ufs_missingop }, 2203 { &vop_reallocblks_desc, (vop_t *) ufs_missingop }, 2204 { &vop_write_desc, (vop_t *) ufs_missingop }, 2205 { &vop_access_desc, (vop_t *) ufs_access }, 2206 { &vop_advlock_desc, (vop_t *) ufs_advlock }, 2207 { &vop_bmap_desc, (vop_t *) ufs_bmap }, 2208 { &vop_cachedlookup_desc, (vop_t *) ufs_lookup }, 2209 { &vop_close_desc, (vop_t *) ufs_close }, 2210 { &vop_create_desc, (vop_t *) ufs_create }, 2211 { &vop_getattr_desc, (vop_t *) ufs_getattr }, 2212 { &vop_inactive_desc, (vop_t *) ufs_inactive }, 2213 { &vop_islocked_desc, (vop_t *) vop_stdislocked }, 2214 { &vop_link_desc, (vop_t *) ufs_link }, 2215 { &vop_lock_desc, (vop_t *) vop_stdlock }, 2216 { &vop_lookup_desc, (vop_t *) vfs_cache_lookup }, 2217 { &vop_mkdir_desc, (vop_t *) ufs_mkdir }, 2218 { &vop_mknod_desc, (vop_t *) ufs_mknod }, 2219 { &vop_mmap_desc, (vop_t *) ufs_mmap }, 2220 { &vop_open_desc, (vop_t *) ufs_open }, 2221 { &vop_pathconf_desc, (vop_t *) ufs_pathconf }, 2222 { &vop_poll_desc, (vop_t *) vop_stdpoll }, 2223 { &vop_getwritemount_desc, (vop_t *) vop_stdgetwritemount }, 2224 { &vop_print_desc, (vop_t *) ufs_print }, 2225 { &vop_readdir_desc, (vop_t *) ufs_readdir }, 2226 { &vop_readlink_desc, (vop_t *) ufs_readlink }, 2227 { &vop_reclaim_desc, (vop_t *) ufs_reclaim }, 2228 { &vop_remove_desc, (vop_t *) ufs_remove }, 2229 { &vop_rename_desc, (vop_t *) ufs_rename }, 2230 { &vop_rmdir_desc, (vop_t *) ufs_rmdir }, 2231 { &vop_setattr_desc, (vop_t *) ufs_setattr }, 2232 { &vop_strategy_desc, (vop_t *) ufs_strategy }, 2233 { &vop_symlink_desc, (vop_t *) ufs_symlink }, 2234 { &vop_unlock_desc, (vop_t *) vop_stdunlock }, 2235 { &vop_whiteout_desc, (vop_t *) ufs_whiteout }, 2236 { NULL, NULL } 2237}; 2238static struct vnodeopv_desc ufs_vnodeop_opv_desc = 2239 { &ufs_vnodeop_p, ufs_vnodeop_entries }; 2240 2241static vop_t **ufs_specop_p; 2242static struct vnodeopv_entry_desc ufs_specop_entries[] = { 2243 { &vop_default_desc, (vop_t *) spec_vnoperate }, 2244 { &vop_fsync_desc, (vop_t *) ufs_missingop }, 2245 { &vop_access_desc, (vop_t *) ufs_access }, 2246 { &vop_close_desc, (vop_t *) ufsspec_close }, 2247 { &vop_getattr_desc, (vop_t *) ufs_getattr }, 2248 { &vop_inactive_desc, (vop_t *) ufs_inactive }, 2249 { &vop_islocked_desc, (vop_t *) vop_stdislocked }, 2250 { &vop_lock_desc, (vop_t *) vop_stdlock }, 2251 { &vop_print_desc, (vop_t *) ufs_print }, 2252 { &vop_read_desc, (vop_t *) ufsspec_read }, 2253 { &vop_reclaim_desc, (vop_t *) ufs_reclaim }, 2254 { &vop_setattr_desc, (vop_t *) ufs_setattr }, 2255 { &vop_unlock_desc, (vop_t *) vop_stdunlock }, 2256 { &vop_write_desc, (vop_t *) ufsspec_write }, 2257 { NULL, NULL } 2258}; 2259static struct vnodeopv_desc ufs_specop_opv_desc = 2260 { &ufs_specop_p, ufs_specop_entries }; 2261 2262static vop_t **ufs_fifoop_p; 2263static struct vnodeopv_entry_desc ufs_fifoop_entries[] = { 2264 { &vop_default_desc, (vop_t *) fifo_vnoperate }, 2265 { &vop_fsync_desc, (vop_t *) ufs_missingop }, 2266 { &vop_access_desc, (vop_t *) ufs_access }, 2267 { &vop_close_desc, (vop_t *) ufsfifo_close }, 2268 { &vop_getattr_desc, (vop_t *) ufs_getattr }, 2269 { &vop_inactive_desc, (vop_t *) ufs_inactive }, 2270 { &vop_islocked_desc, (vop_t *) vop_stdislocked }, 2271 { &vop_lock_desc, (vop_t *) vop_stdlock }, 2272 { &vop_print_desc, (vop_t *) ufs_print }, 2273 { &vop_read_desc, (vop_t *) ufsfifo_read }, 2274 { &vop_reclaim_desc, (vop_t *) ufs_reclaim }, 2275 { &vop_setattr_desc, (vop_t *) ufs_setattr }, 2276 { &vop_unlock_desc, (vop_t *) vop_stdunlock }, 2277 { &vop_write_desc, (vop_t *) ufsfifo_write }, 2278 { NULL, NULL } 2279}; 2280static struct vnodeopv_desc ufs_fifoop_opv_desc = 2281 { &ufs_fifoop_p, ufs_fifoop_entries }; 2282 2283VNODEOP_SET(ufs_vnodeop_opv_desc); 2284VNODEOP_SET(ufs_specop_opv_desc); 2285VNODEOP_SET(ufs_fifoop_opv_desc); 2286 2287int 2288ufs_vnoperate(ap) 2289 struct vop_generic_args /* { 2290 struct vnodeop_desc *a_desc; 2291 } */ *ap; 2292{ 2293 return (VOCALL(ufs_vnodeop_p, ap->a_desc->vdesc_offset, ap)); 2294} 2295 2296int 2297ufs_vnoperatefifo(ap) 2298 struct vop_generic_args /* { 2299 struct vnodeop_desc *a_desc; 2300 } */ *ap; 2301{ 2302 return (VOCALL(ufs_fifoop_p, ap->a_desc->vdesc_offset, ap)); 2303} 2304 2305int 2306ufs_vnoperatespec(ap) 2307 struct vop_generic_args /* { 2308 struct vnodeop_desc *a_desc; 2309 } */ *ap; 2310{ 2311 return (VOCALL(ufs_specop_p, ap->a_desc->vdesc_offset, ap)); 2312} 2313