1 /* $NetBSD: ufs_vnops.c,v 1.169 2008/08/14 16:19:25 matt Exp $ */ 2 3 /*- 4 * Copyright (c) 2008 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Wasabi Systems, 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 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 /* 33 * Copyright (c) 1982, 1986, 1989, 1993, 1995 34 * The Regents of the University of California. All rights reserved. 35 * (c) UNIX System Laboratories, Inc. 36 * All or some portions of this file are derived from material licensed 37 * to the University of California by American Telephone and Telegraph 38 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 39 * the permission of UNIX System Laboratories, Inc. 40 * 41 * Redistribution and use in source and binary forms, with or without 42 * modification, are permitted provided that the following conditions 43 * are met: 44 * 1. Redistributions of source code must retain the above copyright 45 * notice, this list of conditions and the following disclaimer. 46 * 2. Redistributions in binary form must reproduce the above copyright 47 * notice, this list of conditions and the following disclaimer in the 48 * documentation and/or other materials provided with the distribution. 49 * 3. Neither the name of the University nor the names of its contributors 50 * may be used to endorse or promote products derived from this software 51 * without specific prior written permission. 52 * 53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 63 * SUCH DAMAGE. 64 * 65 * @(#)ufs_vnops.c 8.28 (Berkeley) 7/31/95 66 */ 67 68 #include <sys/cdefs.h> 69 __KERNEL_RCSID(0, "$NetBSD: ufs_vnops.c,v 1.169 2008/08/14 16:19:25 matt Exp $"); 70 71 #if defined(_KERNEL_OPT) 72 #include "opt_ffs.h" 73 #include "opt_quota.h" 74 #include "fs_lfs.h" 75 #endif 76 77 #include <sys/param.h> 78 #include <sys/systm.h> 79 #include <sys/namei.h> 80 #include <sys/resourcevar.h> 81 #include <sys/kernel.h> 82 #include <sys/file.h> 83 #include <sys/stat.h> 84 #include <sys/buf.h> 85 #include <sys/proc.h> 86 #include <sys/mount.h> 87 #include <sys/vnode.h> 88 #include <sys/malloc.h> 89 #include <sys/dirent.h> 90 #include <sys/lockf.h> 91 #include <sys/kauth.h> 92 #include <sys/wapbl.h> 93 #include <sys/fstrans.h> 94 95 #include <miscfs/specfs/specdev.h> 96 #include <miscfs/fifofs/fifo.h> 97 98 #include <ufs/ufs/inode.h> 99 #include <ufs/ufs/dir.h> 100 #include <ufs/ufs/ufsmount.h> 101 #include <ufs/ufs/ufs_bswap.h> 102 #include <ufs/ufs/ufs_extern.h> 103 #include <ufs/ufs/ufs_wapbl.h> 104 #ifdef UFS_DIRHASH 105 #include <ufs/ufs/dirhash.h> 106 #endif 107 #include <ufs/ext2fs/ext2fs_extern.h> 108 #include <ufs/ffs/ffs_extern.h> 109 #include <ufs/lfs/lfs_extern.h> 110 111 #include <uvm/uvm.h> 112 113 static int ufs_chmod(struct vnode *, int, kauth_cred_t, struct lwp *); 114 static int ufs_chown(struct vnode *, uid_t, gid_t, kauth_cred_t, 115 struct lwp *); 116 117 /* 118 * A virgin directory (no blushing please). 119 */ 120 static const struct dirtemplate mastertemplate = { 121 0, 12, DT_DIR, 1, ".", 122 0, DIRBLKSIZ - 12, DT_DIR, 2, ".." 123 }; 124 125 /* 126 * Create a regular file 127 */ 128 int 129 ufs_create(void *v) 130 { 131 struct vop_create_args /* { 132 struct vnode *a_dvp; 133 struct vnode **a_vpp; 134 struct componentname *a_cnp; 135 struct vattr *a_vap; 136 } */ *ap = v; 137 int error; 138 139 /* 140 * UFS_WAPBL_BEGIN1(dvp->v_mount, dvp) performed by successful 141 * ufs_makeinode 142 */ 143 fstrans_start(ap->a_dvp->v_mount, FSTRANS_SHARED); 144 error = 145 ufs_makeinode(MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode), 146 ap->a_dvp, ap->a_vpp, ap->a_cnp); 147 if (error) { 148 fstrans_done(ap->a_dvp->v_mount); 149 return (error); 150 } 151 UFS_WAPBL_END1(ap->a_dvp->v_mount, ap->a_dvp); 152 fstrans_done(ap->a_dvp->v_mount); 153 VN_KNOTE(ap->a_dvp, NOTE_WRITE); 154 return (0); 155 } 156 157 /* 158 * Mknod vnode call 159 */ 160 /* ARGSUSED */ 161 int 162 ufs_mknod(void *v) 163 { 164 struct vop_mknod_args /* { 165 struct vnode *a_dvp; 166 struct vnode **a_vpp; 167 struct componentname *a_cnp; 168 struct vattr *a_vap; 169 } */ *ap = v; 170 struct vattr *vap; 171 struct vnode **vpp; 172 struct inode *ip; 173 int error; 174 struct mount *mp; 175 ino_t ino; 176 177 vap = ap->a_vap; 178 vpp = ap->a_vpp; 179 180 /* 181 * UFS_WAPBL_BEGIN1(dvp->v_mount, dvp) performed by successful 182 * ufs_makeinode 183 */ 184 fstrans_start(ap->a_dvp->v_mount, FSTRANS_SHARED); 185 if ((error = 186 ufs_makeinode(MAKEIMODE(vap->va_type, vap->va_mode), 187 ap->a_dvp, vpp, ap->a_cnp)) != 0) 188 goto out; 189 VN_KNOTE(ap->a_dvp, NOTE_WRITE); 190 ip = VTOI(*vpp); 191 mp = (*vpp)->v_mount; 192 ino = ip->i_number; 193 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE; 194 if (vap->va_rdev != VNOVAL) { 195 struct ufsmount *ump = ip->i_ump; 196 /* 197 * Want to be able to use this to make badblock 198 * inodes, so don't truncate the dev number. 199 */ 200 if (ump->um_fstype == UFS1) 201 ip->i_ffs1_rdev = ufs_rw32(vap->va_rdev, 202 UFS_MPNEEDSWAP(ump)); 203 else 204 ip->i_ffs2_rdev = ufs_rw64(vap->va_rdev, 205 UFS_MPNEEDSWAP(ump)); 206 } 207 UFS_WAPBL_UPDATE(*vpp, NULL, NULL, 0); 208 UFS_WAPBL_END1(ap->a_dvp->v_mount, ap->a_dvp); 209 /* 210 * Remove inode so that it will be reloaded by VFS_VGET and 211 * checked to see if it is an alias of an existing entry in 212 * the inode cache. 213 */ 214 VOP_UNLOCK(*vpp, 0); 215 (*vpp)->v_type = VNON; 216 vgone(*vpp); 217 error = VFS_VGET(mp, ino, vpp); 218 out: 219 fstrans_done(ap->a_dvp->v_mount); 220 if (error != 0) { 221 *vpp = NULL; 222 return (error); 223 } 224 return (0); 225 } 226 227 /* 228 * Open called. 229 * 230 * Nothing to do. 231 */ 232 /* ARGSUSED */ 233 int 234 ufs_open(void *v) 235 { 236 struct vop_open_args /* { 237 struct vnode *a_vp; 238 int a_mode; 239 kauth_cred_t a_cred; 240 } */ *ap = v; 241 242 /* 243 * Files marked append-only must be opened for appending. 244 */ 245 if ((VTOI(ap->a_vp)->i_flags & APPEND) && 246 (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE) 247 return (EPERM); 248 return (0); 249 } 250 251 /* 252 * Close called. 253 * 254 * Update the times on the inode. 255 */ 256 /* ARGSUSED */ 257 int 258 ufs_close(void *v) 259 { 260 struct vop_close_args /* { 261 struct vnode *a_vp; 262 int a_fflag; 263 kauth_cred_t a_cred; 264 } */ *ap = v; 265 struct vnode *vp; 266 struct inode *ip; 267 268 vp = ap->a_vp; 269 ip = VTOI(vp); 270 if (vp->v_usecount > 1) 271 UFS_ITIMES(vp, NULL, NULL, NULL); 272 return (0); 273 } 274 275 int 276 ufs_access(void *v) 277 { 278 struct vop_access_args /* { 279 struct vnode *a_vp; 280 int a_mode; 281 kauth_cred_t a_cred; 282 } */ *ap = v; 283 struct vnode *vp; 284 struct inode *ip; 285 mode_t mode; 286 #ifdef QUOTA 287 int error; 288 #endif 289 290 vp = ap->a_vp; 291 ip = VTOI(vp); 292 mode = ap->a_mode; 293 /* 294 * Disallow write attempts on read-only file systems; 295 * unless the file is a socket, fifo, or a block or 296 * character device resident on the file system. 297 */ 298 if (mode & VWRITE) { 299 switch (vp->v_type) { 300 case VDIR: 301 case VLNK: 302 case VREG: 303 if (vp->v_mount->mnt_flag & MNT_RDONLY) 304 return (EROFS); 305 #ifdef QUOTA 306 fstrans_start(vp->v_mount, FSTRANS_SHARED); 307 error = getinoquota(ip); 308 fstrans_done(vp->v_mount); 309 if (error != 0) 310 return error; 311 #endif 312 break; 313 case VBAD: 314 case VBLK: 315 case VCHR: 316 case VSOCK: 317 case VFIFO: 318 case VNON: 319 default: 320 break; 321 } 322 } 323 324 /* If it is a snapshot, nobody gets access to it. */ 325 if ((ip->i_flags & SF_SNAPSHOT)) 326 return (EPERM); 327 /* If immutable bit set, nobody gets to write it. */ 328 if ((mode & VWRITE) && (ip->i_flags & IMMUTABLE)) 329 return (EPERM); 330 331 return (vaccess(vp->v_type, ip->i_mode & ALLPERMS, 332 ip->i_uid, ip->i_gid, mode, ap->a_cred)); 333 } 334 335 /* ARGSUSED */ 336 int 337 ufs_getattr(void *v) 338 { 339 struct vop_getattr_args /* { 340 struct vnode *a_vp; 341 struct vattr *a_vap; 342 kauth_cred_t a_cred; 343 } */ *ap = v; 344 struct vnode *vp; 345 struct inode *ip; 346 struct vattr *vap; 347 348 vp = ap->a_vp; 349 ip = VTOI(vp); 350 vap = ap->a_vap; 351 UFS_ITIMES(vp, NULL, NULL, NULL); 352 353 /* 354 * Copy from inode table 355 */ 356 vap->va_fsid = ip->i_dev; 357 vap->va_fileid = ip->i_number; 358 vap->va_mode = ip->i_mode & ALLPERMS; 359 vap->va_nlink = ip->i_ffs_effnlink; 360 vap->va_uid = ip->i_uid; 361 vap->va_gid = ip->i_gid; 362 vap->va_size = vp->v_size; 363 if (ip->i_ump->um_fstype == UFS1) { 364 vap->va_rdev = (dev_t)ufs_rw32(ip->i_ffs1_rdev, 365 UFS_MPNEEDSWAP(ip->i_ump)); 366 vap->va_atime.tv_sec = ip->i_ffs1_atime; 367 vap->va_atime.tv_nsec = ip->i_ffs1_atimensec; 368 vap->va_mtime.tv_sec = ip->i_ffs1_mtime; 369 vap->va_mtime.tv_nsec = ip->i_ffs1_mtimensec; 370 vap->va_ctime.tv_sec = ip->i_ffs1_ctime; 371 vap->va_ctime.tv_nsec = ip->i_ffs1_ctimensec; 372 vap->va_birthtime.tv_sec = 0; 373 vap->va_birthtime.tv_nsec = 0; 374 vap->va_bytes = dbtob((u_quad_t)ip->i_ffs1_blocks); 375 } else { 376 vap->va_rdev = (dev_t)ufs_rw64(ip->i_ffs2_rdev, 377 UFS_MPNEEDSWAP(ip->i_ump)); 378 vap->va_atime.tv_sec = ip->i_ffs2_atime; 379 vap->va_atime.tv_nsec = ip->i_ffs2_atimensec; 380 vap->va_mtime.tv_sec = ip->i_ffs2_mtime; 381 vap->va_mtime.tv_nsec = ip->i_ffs2_mtimensec; 382 vap->va_ctime.tv_sec = ip->i_ffs2_ctime; 383 vap->va_ctime.tv_nsec = ip->i_ffs2_ctimensec; 384 vap->va_birthtime.tv_sec = ip->i_ffs2_birthtime; 385 vap->va_birthtime.tv_nsec = ip->i_ffs2_birthnsec; 386 vap->va_bytes = dbtob(ip->i_ffs2_blocks); 387 } 388 vap->va_gen = ip->i_gen; 389 vap->va_flags = ip->i_flags; 390 391 /* this doesn't belong here */ 392 if (vp->v_type == VBLK) 393 vap->va_blocksize = BLKDEV_IOSIZE; 394 else if (vp->v_type == VCHR) 395 vap->va_blocksize = MAXBSIZE; 396 else 397 vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize; 398 vap->va_type = vp->v_type; 399 vap->va_filerev = ip->i_modrev; 400 return (0); 401 } 402 403 /* 404 * Set attribute vnode op. called from several syscalls 405 */ 406 int 407 ufs_setattr(void *v) 408 { 409 struct vop_setattr_args /* { 410 struct vnode *a_vp; 411 struct vattr *a_vap; 412 kauth_cred_t a_cred; 413 } */ *ap = v; 414 struct vattr *vap; 415 struct vnode *vp; 416 struct inode *ip; 417 kauth_cred_t cred; 418 struct lwp *l; 419 int error; 420 421 vap = ap->a_vap; 422 vp = ap->a_vp; 423 ip = VTOI(vp); 424 cred = ap->a_cred; 425 l = curlwp; 426 427 /* 428 * Check for unsettable attributes. 429 */ 430 if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) || 431 (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) || 432 (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) || 433 ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) { 434 return (EINVAL); 435 } 436 437 fstrans_start(vp->v_mount, FSTRANS_SHARED); 438 439 if (vap->va_flags != VNOVAL) { 440 if (vp->v_mount->mnt_flag & MNT_RDONLY) { 441 error = EROFS; 442 goto out; 443 } 444 if (kauth_cred_geteuid(cred) != ip->i_uid && 445 (error = kauth_authorize_generic(cred, 446 KAUTH_GENERIC_ISSUSER, NULL))) 447 goto out; 448 if (kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER, 449 NULL) == 0) { 450 if ((ip->i_flags & (SF_IMMUTABLE | SF_APPEND)) && 451 kauth_authorize_system(l->l_cred, 452 KAUTH_SYSTEM_CHSYSFLAGS, 0, NULL, NULL, NULL)) { 453 error = EPERM; 454 goto out; 455 } 456 /* Snapshot flag cannot be set or cleared */ 457 if ((vap->va_flags & SF_SNAPSHOT) != 458 (ip->i_flags & SF_SNAPSHOT)) { 459 error = EPERM; 460 goto out; 461 } 462 error = UFS_WAPBL_BEGIN(vp->v_mount); 463 if (error) 464 goto out; 465 ip->i_flags = vap->va_flags; 466 DIP_ASSIGN(ip, flags, ip->i_flags); 467 } else { 468 if ((ip->i_flags & (SF_IMMUTABLE | SF_APPEND)) || 469 (vap->va_flags & UF_SETTABLE) != vap->va_flags) { 470 error = EPERM; 471 goto out; 472 } 473 if ((ip->i_flags & SF_SETTABLE) != 474 (vap->va_flags & SF_SETTABLE)) { 475 error = EPERM; 476 goto out; 477 } 478 error = UFS_WAPBL_BEGIN(vp->v_mount); 479 if (error) 480 goto out; 481 ip->i_flags &= SF_SETTABLE; 482 ip->i_flags |= (vap->va_flags & UF_SETTABLE); 483 DIP_ASSIGN(ip, flags, ip->i_flags); 484 } 485 ip->i_flag |= IN_CHANGE; 486 UFS_WAPBL_UPDATE(vp, NULL, NULL, 0); 487 UFS_WAPBL_END(vp->v_mount); 488 if (vap->va_flags & (IMMUTABLE | APPEND)) { 489 error = 0; 490 goto out; 491 } 492 } 493 if (ip->i_flags & (IMMUTABLE | APPEND)) { 494 error = EPERM; 495 goto out; 496 } 497 /* 498 * Go through the fields and update iff not VNOVAL. 499 */ 500 if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) { 501 if (vp->v_mount->mnt_flag & MNT_RDONLY) { 502 error = EROFS; 503 goto out; 504 } 505 error = UFS_WAPBL_BEGIN(vp->v_mount); 506 if (error) 507 goto out; 508 error = ufs_chown(vp, vap->va_uid, vap->va_gid, cred, l); 509 UFS_WAPBL_END(vp->v_mount); 510 if (error) 511 goto out; 512 } 513 if (vap->va_size != VNOVAL) { 514 /* 515 * Disallow write attempts on read-only file systems; 516 * unless the file is a socket, fifo, or a block or 517 * character device resident on the file system. 518 */ 519 switch (vp->v_type) { 520 case VDIR: 521 error = EISDIR; 522 goto out; 523 case VCHR: 524 case VBLK: 525 case VFIFO: 526 break; 527 case VREG: 528 if (vp->v_mount->mnt_flag & MNT_RDONLY) { 529 error = EROFS; 530 goto out; 531 } 532 if ((ip->i_flags & SF_SNAPSHOT) != 0) { 533 error = EPERM; 534 goto out; 535 } 536 error = UFS_WAPBL_BEGIN(vp->v_mount); 537 if (error) 538 goto out; 539 /* 540 * When journaling, only truncate one indirect block 541 * at a time. 542 */ 543 if (vp->v_mount->mnt_wapbl) { 544 uint64_t incr = MNINDIR(ip->i_ump) << 545 vp->v_mount->mnt_fs_bshift; /* Power of 2 */ 546 uint64_t base = NDADDR << 547 vp->v_mount->mnt_fs_bshift; 548 while (!error && ip->i_size > base + incr && 549 ip->i_size > vap->va_size + incr) { 550 /* 551 * round down to next full indirect 552 * block boundary. 553 */ 554 uint64_t nsize = base + 555 ((ip->i_size - base - 1) & 556 ~(incr - 1)); 557 error = UFS_TRUNCATE(vp, nsize, 0, 558 cred); 559 if (error == 0) { 560 UFS_WAPBL_END(vp->v_mount); 561 error = 562 UFS_WAPBL_BEGIN(vp->v_mount); 563 } 564 } 565 } 566 if (!error) 567 error = UFS_TRUNCATE(vp, vap->va_size, 0, cred); 568 UFS_WAPBL_END(vp->v_mount); 569 if (error) 570 goto out; 571 break; 572 default: 573 error = EOPNOTSUPP; 574 goto out; 575 } 576 } 577 ip = VTOI(vp); 578 if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL || 579 vap->va_birthtime.tv_sec != VNOVAL) { 580 if (vp->v_mount->mnt_flag & MNT_RDONLY) { 581 error = EROFS; 582 goto out; 583 } 584 if ((ip->i_flags & SF_SNAPSHOT) != 0) { 585 error = EPERM; 586 goto out; 587 } 588 if (kauth_cred_geteuid(cred) != ip->i_uid && 589 (error = kauth_authorize_generic(cred, 590 KAUTH_GENERIC_ISSUSER, NULL)) && 591 ((vap->va_vaflags & VA_UTIMES_NULL) == 0 || 592 (error = VOP_ACCESS(vp, VWRITE, cred)))) 593 goto out; 594 error = UFS_WAPBL_BEGIN(vp->v_mount); 595 if (error) 596 goto out; 597 if (vap->va_atime.tv_sec != VNOVAL) 598 if (!(vp->v_mount->mnt_flag & MNT_NOATIME)) 599 ip->i_flag |= IN_ACCESS; 600 if (vap->va_mtime.tv_sec != VNOVAL) 601 ip->i_flag |= IN_CHANGE | IN_UPDATE; 602 if (vap->va_birthtime.tv_sec != VNOVAL && 603 ip->i_ump->um_fstype == UFS2) { 604 ip->i_ffs2_birthtime = vap->va_birthtime.tv_sec; 605 ip->i_ffs2_birthnsec = vap->va_birthtime.tv_nsec; 606 } 607 error = UFS_UPDATE(vp, &vap->va_atime, &vap->va_mtime, 0); 608 UFS_WAPBL_END(vp->v_mount); 609 if (error) 610 goto out; 611 } 612 error = 0; 613 if (vap->va_mode != (mode_t)VNOVAL) { 614 if (vp->v_mount->mnt_flag & MNT_RDONLY) { 615 error = EROFS; 616 goto out; 617 } 618 if ((ip->i_flags & SF_SNAPSHOT) != 0 && 619 (vap->va_mode & (S_IXUSR | S_IWUSR | S_IXGRP | S_IWGRP | 620 S_IXOTH | S_IWOTH))) { 621 error = EPERM; 622 goto out; 623 } 624 error = UFS_WAPBL_BEGIN(vp->v_mount); 625 if (error) 626 goto out; 627 error = ufs_chmod(vp, (int)vap->va_mode, cred, l); 628 UFS_WAPBL_END(vp->v_mount); 629 } 630 VN_KNOTE(vp, NOTE_ATTRIB); 631 out: 632 fstrans_done(vp->v_mount); 633 return (error); 634 } 635 636 /* 637 * Change the mode on a file. 638 * Inode must be locked before calling. 639 */ 640 static int 641 ufs_chmod(struct vnode *vp, int mode, kauth_cred_t cred, struct lwp *l) 642 { 643 struct inode *ip; 644 int error, ismember = 0; 645 646 UFS_WAPBL_JLOCK_ASSERT(vp->v_mount); 647 648 ip = VTOI(vp); 649 if (kauth_cred_geteuid(cred) != ip->i_uid && 650 (error = kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER, NULL))) 651 return (error); 652 if (kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER, NULL)) { 653 if (vp->v_type != VDIR && (mode & S_ISTXT)) 654 return (EFTYPE); 655 if ((kauth_cred_ismember_gid(cred, ip->i_gid, &ismember) != 0 || 656 !ismember) && (mode & ISGID)) 657 return (EPERM); 658 } 659 ip->i_mode &= ~ALLPERMS; 660 ip->i_mode |= (mode & ALLPERMS); 661 ip->i_flag |= IN_CHANGE; 662 DIP_ASSIGN(ip, mode, ip->i_mode); 663 UFS_WAPBL_UPDATE(vp, NULL, NULL, 0); 664 return (0); 665 } 666 667 /* 668 * Perform chown operation on inode ip; 669 * inode must be locked prior to call. 670 */ 671 static int 672 ufs_chown(struct vnode *vp, uid_t uid, gid_t gid, kauth_cred_t cred, 673 struct lwp *l) 674 { 675 struct inode *ip; 676 int error, ismember = 0; 677 #ifdef QUOTA 678 uid_t ouid; 679 gid_t ogid; 680 int64_t change; 681 #endif 682 ip = VTOI(vp); 683 error = 0; 684 685 if (uid == (uid_t)VNOVAL) 686 uid = ip->i_uid; 687 if (gid == (gid_t)VNOVAL) 688 gid = ip->i_gid; 689 /* 690 * If we don't own the file, are trying to change the owner 691 * of the file, or are not a member of the target group, 692 * the caller's credentials must imply super-user privilege 693 * or the call fails. 694 */ 695 if ((kauth_cred_geteuid(cred) != ip->i_uid || uid != ip->i_uid || 696 (gid != ip->i_gid && 697 !(kauth_cred_getegid(cred) == gid || 698 (kauth_cred_ismember_gid(cred, gid, &ismember) == 0 && 699 ismember)))) && 700 ((error = kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER, 701 NULL)) != 0)) 702 return (error); 703 704 #ifdef QUOTA 705 ogid = ip->i_gid; 706 ouid = ip->i_uid; 707 change = DIP(ip, blocks); 708 (void) chkdq(ip, -change, cred, 0); 709 (void) chkiq(ip, -1, cred, 0); 710 #endif 711 ip->i_gid = gid; 712 DIP_ASSIGN(ip, gid, gid); 713 ip->i_uid = uid; 714 DIP_ASSIGN(ip, uid, uid); 715 #ifdef QUOTA 716 if ((error = chkdq(ip, change, cred, 0)) == 0) { 717 if ((error = chkiq(ip, 1, cred, 0)) == 0) 718 goto good; 719 else 720 (void) chkdq(ip, -change, cred, FORCE); 721 } 722 ip->i_gid = ogid; 723 DIP_ASSIGN(ip, gid, ogid); 724 ip->i_uid = ouid; 725 DIP_ASSIGN(ip, uid, ouid); 726 (void) chkdq(ip, change, cred, FORCE); 727 (void) chkiq(ip, 1, cred, FORCE); 728 return (error); 729 good: 730 #endif /* QUOTA */ 731 ip->i_flag |= IN_CHANGE; 732 UFS_WAPBL_UPDATE(vp, NULL, NULL, 0); 733 return (0); 734 } 735 736 int 737 ufs_remove(void *v) 738 { 739 struct vop_remove_args /* { 740 struct vnode *a_dvp; 741 struct vnode *a_vp; 742 struct componentname *a_cnp; 743 } */ *ap = v; 744 struct vnode *vp, *dvp; 745 struct inode *ip; 746 int error; 747 bool pace; 748 749 vp = ap->a_vp; 750 dvp = ap->a_dvp; 751 ip = VTOI(vp); 752 fstrans_start(dvp->v_mount, FSTRANS_SHARED); 753 if (vp->v_type == VDIR || (ip->i_flags & (IMMUTABLE | APPEND)) || 754 (VTOI(dvp)->i_flags & APPEND)) 755 error = EPERM; 756 else { 757 error = UFS_WAPBL_BEGIN(dvp->v_mount); 758 if (error == 0) { 759 error = ufs_dirremove(dvp, ip, ap->a_cnp->cn_flags, 0); 760 UFS_WAPBL_END(dvp->v_mount); 761 } 762 } 763 VN_KNOTE(vp, NOTE_DELETE); 764 VN_KNOTE(dvp, NOTE_WRITE); 765 if (dvp == vp) 766 vrele(vp); 767 else 768 vput(vp); 769 pace = DOINGSOFTDEP(dvp); 770 vput(dvp); 771 fstrans_done(dvp->v_mount); 772 if (pace) { 773 /* 774 * Give the syncer some breathing room so that it 775 * can flush removes. XXX 776 */ 777 softdep_pace_dirrem(); 778 } 779 return (error); 780 } 781 782 /* 783 * link vnode call 784 */ 785 int 786 ufs_link(void *v) 787 { 788 struct vop_link_args /* { 789 struct vnode *a_dvp; 790 struct vnode *a_vp; 791 struct componentname *a_cnp; 792 } */ *ap = v; 793 struct vnode *vp, *dvp; 794 struct componentname *cnp; 795 struct inode *ip; 796 struct direct *newdir; 797 int error; 798 799 dvp = ap->a_dvp; 800 vp = ap->a_vp; 801 cnp = ap->a_cnp; 802 #ifdef DIAGNOSTIC 803 if ((cnp->cn_flags & HASBUF) == 0) 804 panic("ufs_link: no name"); 805 #endif 806 fstrans_start(dvp->v_mount, FSTRANS_SHARED); 807 if (vp->v_type == VDIR) { 808 VOP_ABORTOP(dvp, cnp); 809 error = EPERM; 810 goto out2; 811 } 812 if (dvp->v_mount != vp->v_mount) { 813 VOP_ABORTOP(dvp, cnp); 814 error = EXDEV; 815 goto out2; 816 } 817 if (dvp != vp && (error = vn_lock(vp, LK_EXCLUSIVE))) { 818 VOP_ABORTOP(dvp, cnp); 819 goto out2; 820 } 821 ip = VTOI(vp); 822 if ((nlink_t)ip->i_nlink >= LINK_MAX) { 823 VOP_ABORTOP(dvp, cnp); 824 error = EMLINK; 825 goto out1; 826 } 827 if (ip->i_flags & (IMMUTABLE | APPEND)) { 828 VOP_ABORTOP(dvp, cnp); 829 error = EPERM; 830 goto out1; 831 } 832 error = UFS_WAPBL_BEGIN(vp->v_mount); 833 if (error) { 834 VOP_ABORTOP(dvp, cnp); 835 goto out1; 836 } 837 ip->i_ffs_effnlink++; 838 ip->i_nlink++; 839 DIP_ASSIGN(ip, nlink, ip->i_nlink); 840 ip->i_flag |= IN_CHANGE; 841 if (DOINGSOFTDEP(vp)) 842 softdep_change_linkcnt(ip); 843 error = UFS_UPDATE(vp, NULL, NULL, UPDATE_DIROP); 844 if (!error) { 845 newdir = pool_cache_get(ufs_direct_cache, PR_WAITOK); 846 ufs_makedirentry(ip, cnp, newdir); 847 error = ufs_direnter(dvp, vp, newdir, cnp, NULL); 848 pool_cache_put(ufs_direct_cache, newdir); 849 } 850 if (error) { 851 ip->i_ffs_effnlink--; 852 ip->i_nlink--; 853 DIP_ASSIGN(ip, nlink, ip->i_nlink); 854 ip->i_flag |= IN_CHANGE; 855 UFS_WAPBL_UPDATE(vp, NULL, NULL, UPDATE_DIROP); 856 if (DOINGSOFTDEP(vp)) 857 softdep_change_linkcnt(ip); 858 } 859 PNBUF_PUT(cnp->cn_pnbuf); 860 UFS_WAPBL_END(vp->v_mount); 861 out1: 862 if (dvp != vp) 863 VOP_UNLOCK(vp, 0); 864 out2: 865 VN_KNOTE(vp, NOTE_LINK); 866 VN_KNOTE(dvp, NOTE_WRITE); 867 vput(dvp); 868 fstrans_done(dvp->v_mount); 869 return (error); 870 } 871 872 /* 873 * whiteout vnode call 874 */ 875 int 876 ufs_whiteout(void *v) 877 { 878 struct vop_whiteout_args /* { 879 struct vnode *a_dvp; 880 struct componentname *a_cnp; 881 int a_flags; 882 } */ *ap = v; 883 struct vnode *dvp = ap->a_dvp; 884 struct componentname *cnp = ap->a_cnp; 885 struct direct *newdir; 886 int error; 887 struct ufsmount *ump = VFSTOUFS(dvp->v_mount); 888 889 error = 0; 890 switch (ap->a_flags) { 891 case LOOKUP: 892 /* 4.4 format directories support whiteout operations */ 893 if (ump->um_maxsymlinklen > 0) 894 return (0); 895 return (EOPNOTSUPP); 896 897 case CREATE: 898 /* create a new directory whiteout */ 899 fstrans_start(dvp->v_mount, FSTRANS_SHARED); 900 #ifdef DIAGNOSTIC 901 if ((cnp->cn_flags & SAVENAME) == 0) 902 panic("ufs_whiteout: missing name"); 903 if (ump->um_maxsymlinklen <= 0) 904 panic("ufs_whiteout: old format filesystem"); 905 #endif 906 907 newdir = pool_cache_get(ufs_direct_cache, PR_WAITOK); 908 newdir->d_ino = WINO; 909 newdir->d_namlen = cnp->cn_namelen; 910 memcpy(newdir->d_name, cnp->cn_nameptr, 911 (size_t)cnp->cn_namelen); 912 newdir->d_name[cnp->cn_namelen] = '\0'; 913 newdir->d_type = DT_WHT; 914 error = ufs_direnter(dvp, NULL, newdir, cnp, NULL); 915 pool_cache_put(ufs_direct_cache, newdir); 916 break; 917 918 case DELETE: 919 /* remove an existing directory whiteout */ 920 fstrans_start(dvp->v_mount, FSTRANS_SHARED); 921 #ifdef DIAGNOSTIC 922 if (ump->um_maxsymlinklen <= 0) 923 panic("ufs_whiteout: old format filesystem"); 924 #endif 925 926 cnp->cn_flags &= ~DOWHITEOUT; 927 error = ufs_dirremove(dvp, NULL, cnp->cn_flags, 0); 928 break; 929 default: 930 panic("ufs_whiteout: unknown op"); 931 /* NOTREACHED */ 932 } 933 if (cnp->cn_flags & HASBUF) { 934 PNBUF_PUT(cnp->cn_pnbuf); 935 cnp->cn_flags &= ~HASBUF; 936 } 937 fstrans_done(dvp->v_mount); 938 return (error); 939 } 940 941 942 /* 943 * Rename vnode operation 944 * rename("foo", "bar"); 945 * is essentially 946 * unlink("bar"); 947 * link("foo", "bar"); 948 * unlink("foo"); 949 * but ``atomically''. Can't do full commit without saving state in the 950 * inode on disk which isn't feasible at this time. Best we can do is 951 * always guarantee the target exists. 952 * 953 * Basic algorithm is: 954 * 955 * 1) Bump link count on source while we're linking it to the 956 * target. This also ensure the inode won't be deleted out 957 * from underneath us while we work (it may be truncated by 958 * a concurrent `trunc' or `open' for creation). 959 * 2) Link source to destination. If destination already exists, 960 * delete it first. 961 * 3) Unlink source reference to inode if still around. If a 962 * directory was moved and the parent of the destination 963 * is different from the source, patch the ".." entry in the 964 * directory. 965 */ 966 int 967 ufs_rename(void *v) 968 { 969 struct vop_rename_args /* { 970 struct vnode *a_fdvp; 971 struct vnode *a_fvp; 972 struct componentname *a_fcnp; 973 struct vnode *a_tdvp; 974 struct vnode *a_tvp; 975 struct componentname *a_tcnp; 976 } */ *ap = v; 977 struct vnode *tvp, *tdvp, *fvp, *fdvp; 978 struct componentname *tcnp, *fcnp; 979 struct inode *ip, *xp, *dp; 980 struct mount *mp; 981 struct direct *newdir; 982 int doingdirectory, oldparent, newparent, error; 983 984 #ifdef WAPBL 985 if (ap->a_tdvp->v_mount->mnt_wapbl) 986 return wapbl_ufs_rename(v); 987 #endif 988 989 tvp = ap->a_tvp; 990 tdvp = ap->a_tdvp; 991 fvp = ap->a_fvp; 992 fdvp = ap->a_fdvp; 993 tcnp = ap->a_tcnp; 994 fcnp = ap->a_fcnp; 995 doingdirectory = oldparent = newparent = error = 0; 996 997 #ifdef DIAGNOSTIC 998 if ((tcnp->cn_flags & HASBUF) == 0 || 999 (fcnp->cn_flags & HASBUF) == 0) 1000 panic("ufs_rename: no name"); 1001 #endif 1002 /* 1003 * Check for cross-device rename. 1004 */ 1005 if ((fvp->v_mount != tdvp->v_mount) || 1006 (tvp && (fvp->v_mount != tvp->v_mount))) { 1007 error = EXDEV; 1008 abortit: 1009 VOP_ABORTOP(tdvp, tcnp); /* XXX, why not in NFS? */ 1010 if (tdvp == tvp) 1011 vrele(tdvp); 1012 else 1013 vput(tdvp); 1014 if (tvp) 1015 vput(tvp); 1016 VOP_ABORTOP(fdvp, fcnp); /* XXX, why not in NFS? */ 1017 vrele(fdvp); 1018 vrele(fvp); 1019 return (error); 1020 } 1021 1022 /* 1023 * Check if just deleting a link name. 1024 */ 1025 if (tvp && ((VTOI(tvp)->i_flags & (IMMUTABLE | APPEND)) || 1026 (VTOI(tdvp)->i_flags & APPEND))) { 1027 error = EPERM; 1028 goto abortit; 1029 } 1030 if (fvp == tvp) { 1031 if (fvp->v_type == VDIR) { 1032 error = EINVAL; 1033 goto abortit; 1034 } 1035 1036 /* Release destination completely. */ 1037 VOP_ABORTOP(tdvp, tcnp); 1038 vput(tdvp); 1039 vput(tvp); 1040 1041 /* Delete source. */ 1042 vrele(fvp); 1043 fcnp->cn_flags &= ~(MODMASK | SAVESTART); 1044 fcnp->cn_flags |= LOCKPARENT | LOCKLEAF; 1045 fcnp->cn_nameiop = DELETE; 1046 vn_lock(fdvp, LK_EXCLUSIVE | LK_RETRY); 1047 if ((error = relookup(fdvp, &fvp, fcnp))) { 1048 vput(fdvp); 1049 return (error); 1050 } 1051 return (VOP_REMOVE(fdvp, fvp, fcnp)); 1052 } 1053 if ((error = vn_lock(fvp, LK_EXCLUSIVE)) != 0) 1054 goto abortit; 1055 dp = VTOI(fdvp); 1056 ip = VTOI(fvp); 1057 if ((nlink_t) ip->i_nlink >= LINK_MAX) { 1058 VOP_UNLOCK(fvp, 0); 1059 error = EMLINK; 1060 goto abortit; 1061 } 1062 if ((ip->i_flags & (IMMUTABLE | APPEND)) || 1063 (dp->i_flags & APPEND)) { 1064 VOP_UNLOCK(fvp, 0); 1065 error = EPERM; 1066 goto abortit; 1067 } 1068 if ((ip->i_mode & IFMT) == IFDIR) { 1069 /* 1070 * Avoid ".", "..", and aliases of "." for obvious reasons. 1071 */ 1072 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') || 1073 dp == ip || 1074 (fcnp->cn_flags & ISDOTDOT) || 1075 (tcnp->cn_flags & ISDOTDOT) || 1076 (ip->i_flag & IN_RENAME)) { 1077 VOP_UNLOCK(fvp, 0); 1078 error = EINVAL; 1079 goto abortit; 1080 } 1081 ip->i_flag |= IN_RENAME; 1082 oldparent = dp->i_number; 1083 doingdirectory = 1; 1084 } 1085 VN_KNOTE(fdvp, NOTE_WRITE); /* XXXLUKEM/XXX: right place? */ 1086 1087 /* 1088 * When the target exists, both the directory 1089 * and target vnodes are returned locked. 1090 */ 1091 dp = VTOI(tdvp); 1092 xp = NULL; 1093 if (tvp) 1094 xp = VTOI(tvp); 1095 1096 mp = fdvp->v_mount; 1097 fstrans_start(mp, FSTRANS_SHARED); 1098 1099 /* 1100 * 1) Bump link count while we're moving stuff 1101 * around. If we crash somewhere before 1102 * completing our work, the link count 1103 * may be wrong, but correctable. 1104 */ 1105 ip->i_ffs_effnlink++; 1106 ip->i_nlink++; 1107 DIP_ASSIGN(ip, nlink, ip->i_nlink); 1108 ip->i_flag |= IN_CHANGE; 1109 if (DOINGSOFTDEP(fvp)) 1110 softdep_change_linkcnt(ip); 1111 if ((error = UFS_UPDATE(fvp, NULL, NULL, UPDATE_DIROP)) != 0) { 1112 VOP_UNLOCK(fvp, 0); 1113 goto bad; 1114 } 1115 1116 /* 1117 * If ".." must be changed (ie the directory gets a new 1118 * parent) then the source directory must not be in the 1119 * directory hierarchy above the target, as this would 1120 * orphan everything below the source directory. Also 1121 * the user must have write permission in the source so 1122 * as to be able to change "..". We must repeat the call 1123 * to namei, as the parent directory is unlocked by the 1124 * call to checkpath(). 1125 */ 1126 error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred); 1127 VOP_UNLOCK(fvp, 0); 1128 if (oldparent != dp->i_number) 1129 newparent = dp->i_number; 1130 if (doingdirectory && newparent) { 1131 if (error) /* write access check above */ 1132 goto bad; 1133 if (xp != NULL) 1134 vput(tvp); 1135 vref(tdvp); /* compensate for the ref checkpath loses */ 1136 if ((error = ufs_checkpath(ip, dp, tcnp->cn_cred)) != 0) { 1137 vrele(tdvp); 1138 goto out; 1139 } 1140 tcnp->cn_flags &= ~SAVESTART; 1141 vn_lock(tdvp, LK_EXCLUSIVE | LK_RETRY); 1142 error = relookup(tdvp, &tvp, tcnp); 1143 if (error != 0) { 1144 vput(tdvp); 1145 goto out; 1146 } 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("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_ffs_effnlink++; 1173 dp->i_nlink++; 1174 DIP_ASSIGN(dp, nlink, dp->i_nlink); 1175 dp->i_flag |= IN_CHANGE; 1176 if (DOINGSOFTDEP(tdvp)) 1177 softdep_change_linkcnt(dp); 1178 if ((error = UFS_UPDATE(tdvp, NULL, NULL, 1179 UPDATE_DIROP)) != 0) { 1180 dp->i_ffs_effnlink--; 1181 dp->i_nlink--; 1182 DIP_ASSIGN(dp, nlink, dp->i_nlink); 1183 dp->i_flag |= IN_CHANGE; 1184 if (DOINGSOFTDEP(tdvp)) 1185 softdep_change_linkcnt(dp); 1186 goto bad; 1187 } 1188 } 1189 newdir = pool_cache_get(ufs_direct_cache, PR_WAITOK); 1190 ufs_makedirentry(ip, tcnp, newdir); 1191 error = ufs_direnter(tdvp, NULL, newdir, tcnp, NULL); 1192 pool_cache_put(ufs_direct_cache, newdir); 1193 if (error != 0) { 1194 if (doingdirectory && newparent) { 1195 dp->i_ffs_effnlink--; 1196 dp->i_nlink--; 1197 DIP_ASSIGN(dp, nlink, dp->i_nlink); 1198 dp->i_flag |= IN_CHANGE; 1199 if (DOINGSOFTDEP(tdvp)) 1200 softdep_change_linkcnt(dp); 1201 (void)UFS_UPDATE(tdvp, NULL, NULL, 1202 UPDATE_WAIT|UPDATE_DIROP); 1203 } 1204 goto bad; 1205 } 1206 VN_KNOTE(tdvp, NOTE_WRITE); 1207 vput(tdvp); 1208 } else { 1209 if (xp->i_dev != dp->i_dev || xp->i_dev != ip->i_dev) 1210 panic("rename: EXDEV"); 1211 /* 1212 * Short circuit rename(foo, foo). 1213 */ 1214 if (xp->i_number == ip->i_number) 1215 panic("rename: same file"); 1216 /* 1217 * If the parent directory is "sticky", then the user must 1218 * own the parent directory, or the destination of the rename, 1219 * otherwise the destination may not be changed (except by 1220 * root). This implements append-only directories. 1221 */ 1222 if ((dp->i_mode & S_ISTXT) && 1223 kauth_authorize_generic(tcnp->cn_cred, 1224 KAUTH_GENERIC_ISSUSER, NULL) != 0 && 1225 kauth_cred_geteuid(tcnp->cn_cred) != dp->i_uid && 1226 xp->i_uid != kauth_cred_geteuid(tcnp->cn_cred)) { 1227 error = EPERM; 1228 goto bad; 1229 } 1230 /* 1231 * Target must be empty if a directory and have no links 1232 * to it. Also, ensure source and target are compatible 1233 * (both directories, or both not directories). 1234 */ 1235 if ((xp->i_mode & IFMT) == IFDIR) { 1236 if (xp->i_ffs_effnlink > 2 || 1237 !ufs_dirempty(xp, dp->i_number, tcnp->cn_cred)) { 1238 error = ENOTEMPTY; 1239 goto bad; 1240 } 1241 if (!doingdirectory) { 1242 error = ENOTDIR; 1243 goto bad; 1244 } 1245 cache_purge(tdvp); 1246 } else if (doingdirectory) { 1247 error = EISDIR; 1248 goto bad; 1249 } 1250 if ((error = ufs_dirrewrite(dp, xp, ip->i_number, 1251 IFTODT(ip->i_mode), doingdirectory && newparent ? 1252 newparent : doingdirectory, IN_CHANGE | IN_UPDATE)) != 0) 1253 goto bad; 1254 if (doingdirectory) { 1255 if (!newparent) { 1256 dp->i_ffs_effnlink--; 1257 if (DOINGSOFTDEP(tdvp)) 1258 softdep_change_linkcnt(dp); 1259 } 1260 xp->i_ffs_effnlink--; 1261 if (DOINGSOFTDEP(tvp)) 1262 softdep_change_linkcnt(xp); 1263 } 1264 if (doingdirectory && !DOINGSOFTDEP(tvp)) { 1265 /* 1266 * Truncate inode. The only stuff left in the directory 1267 * is "." and "..". The "." reference is inconsequential 1268 * since we are quashing it. We have removed the "." 1269 * reference and the reference in the parent directory, 1270 * but there may be other hard links. The soft 1271 * dependency code will arrange to do these operations 1272 * after the parent directory entry has been deleted on 1273 * disk, so when running with that code we avoid doing 1274 * them now. 1275 */ 1276 if (!newparent) { 1277 dp->i_nlink--; 1278 DIP_ASSIGN(dp, nlink, dp->i_nlink); 1279 dp->i_flag |= IN_CHANGE; 1280 } 1281 xp->i_nlink--; 1282 DIP_ASSIGN(xp, nlink, xp->i_nlink); 1283 xp->i_flag |= IN_CHANGE; 1284 if ((error = UFS_TRUNCATE(tvp, (off_t)0, IO_SYNC, 1285 tcnp->cn_cred))) 1286 goto bad; 1287 } 1288 VN_KNOTE(tdvp, NOTE_WRITE); 1289 vput(tdvp); 1290 VN_KNOTE(tvp, NOTE_DELETE); 1291 vput(tvp); 1292 xp = NULL; 1293 } 1294 1295 /* 1296 * 3) Unlink the source. 1297 */ 1298 fcnp->cn_flags &= ~(MODMASK | SAVESTART); 1299 fcnp->cn_flags |= LOCKPARENT | LOCKLEAF; 1300 vn_lock(fdvp, LK_EXCLUSIVE | LK_RETRY); 1301 if ((error = relookup(fdvp, &fvp, fcnp))) { 1302 vput(fdvp); 1303 vrele(ap->a_fvp); 1304 goto out2; 1305 } 1306 if (fvp != NULL) { 1307 xp = VTOI(fvp); 1308 dp = VTOI(fdvp); 1309 } else { 1310 /* 1311 * From name has disappeared. 1312 */ 1313 if (doingdirectory) 1314 panic("rename: lost dir entry"); 1315 vrele(ap->a_fvp); 1316 error = 0; 1317 goto out2; 1318 } 1319 /* 1320 * Ensure that the directory entry still exists and has not 1321 * changed while the new name has been entered. If the source is 1322 * a file then the entry may have been unlinked or renamed. In 1323 * either case there is no further work to be done. If the source 1324 * is a directory then it cannot have been rmdir'ed; The IRENAME 1325 * flag ensures that it cannot be moved by another rename or removed 1326 * by a rmdir. 1327 */ 1328 if (xp != ip) { 1329 if (doingdirectory) 1330 panic("rename: lost dir entry"); 1331 } else { 1332 /* 1333 * If the source is a directory with a 1334 * new parent, the link count of the old 1335 * parent directory must be decremented 1336 * and ".." set to point to the new parent. 1337 */ 1338 if (doingdirectory && newparent) { 1339 KASSERT(dp != NULL); 1340 xp->i_offset = mastertemplate.dot_reclen; 1341 ufs_dirrewrite(xp, dp, newparent, DT_DIR, 0, IN_CHANGE); 1342 cache_purge(fdvp); 1343 } 1344 error = ufs_dirremove(fdvp, xp, fcnp->cn_flags, 0); 1345 xp->i_flag &= ~IN_RENAME; 1346 } 1347 VN_KNOTE(fvp, NOTE_RENAME); 1348 if (dp) 1349 vput(fdvp); 1350 if (xp) 1351 vput(fvp); 1352 vrele(ap->a_fvp); 1353 goto out2; 1354 1355 /* exit routines from steps 1 & 2 */ 1356 bad: 1357 if (xp) 1358 vput(ITOV(xp)); 1359 vput(ITOV(dp)); 1360 out: 1361 if (doingdirectory) 1362 ip->i_flag &= ~IN_RENAME; 1363 if (vn_lock(fvp, LK_EXCLUSIVE) == 0) { 1364 ip->i_ffs_effnlink--; 1365 ip->i_nlink--; 1366 DIP_ASSIGN(ip, nlink, ip->i_nlink); 1367 ip->i_flag |= IN_CHANGE; 1368 ip->i_flag &= ~IN_RENAME; 1369 if (DOINGSOFTDEP(fvp)) 1370 softdep_change_linkcnt(ip); 1371 vput(fvp); 1372 } else 1373 vrele(fvp); 1374 vrele(fdvp); 1375 1376 /* exit routines from step 3 */ 1377 out2: 1378 fstrans_done(mp); 1379 return (error); 1380 } 1381 1382 int 1383 ufs_mkdir(void *v) 1384 { 1385 struct vop_mkdir_args /* { 1386 struct vnode *a_dvp; 1387 struct vnode **a_vpp; 1388 struct componentname *a_cnp; 1389 struct vattr *a_vap; 1390 } */ *ap = v; 1391 struct vnode *dvp = ap->a_dvp, *tvp; 1392 struct vattr *vap = ap->a_vap; 1393 struct componentname *cnp = ap->a_cnp; 1394 struct inode *ip, *dp = VTOI(dvp); 1395 struct buf *bp; 1396 struct dirtemplate dirtemplate; 1397 struct direct *newdir; 1398 int error, dmode, blkoff; 1399 struct ufsmount *ump = dp->i_ump; 1400 int dirblksiz = ump->um_dirblksiz; 1401 1402 fstrans_start(dvp->v_mount, FSTRANS_SHARED); 1403 1404 #ifdef DIAGNOSTIC 1405 if ((cnp->cn_flags & HASBUF) == 0) 1406 panic("ufs_mkdir: no name"); 1407 #endif 1408 if ((nlink_t)dp->i_nlink >= LINK_MAX) { 1409 error = EMLINK; 1410 goto out; 1411 } 1412 dmode = vap->va_mode & ACCESSPERMS; 1413 dmode |= IFDIR; 1414 /* 1415 * Must simulate part of ufs_makeinode here to acquire the inode, 1416 * but not have it entered in the parent directory. The entry is 1417 * made later after writing "." and ".." entries. 1418 */ 1419 if ((error = UFS_VALLOC(dvp, dmode, cnp->cn_cred, ap->a_vpp)) != 0) 1420 goto out; 1421 error = UFS_WAPBL_BEGIN(ap->a_dvp->v_mount); 1422 if (error) 1423 goto out; 1424 tvp = *ap->a_vpp; 1425 ip = VTOI(tvp); 1426 ip->i_uid = kauth_cred_geteuid(cnp->cn_cred); 1427 DIP_ASSIGN(ip, uid, ip->i_uid); 1428 ip->i_gid = dp->i_gid; 1429 DIP_ASSIGN(ip, gid, ip->i_gid); 1430 #ifdef QUOTA 1431 if ((error = chkiq(ip, 1, cnp->cn_cred, 0))) { 1432 PNBUF_PUT(cnp->cn_pnbuf); 1433 UFS_VFREE(tvp, ip->i_number, dmode); 1434 UFS_WAPBL_END(dvp->v_mount); 1435 fstrans_done(dvp->v_mount); 1436 vput(tvp); 1437 vput(dvp); 1438 return (error); 1439 } 1440 #endif 1441 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE; 1442 ip->i_mode = dmode; 1443 DIP_ASSIGN(ip, mode, dmode); 1444 tvp->v_type = VDIR; /* Rest init'd in getnewvnode(). */ 1445 ip->i_ffs_effnlink = 2; 1446 ip->i_nlink = 2; 1447 DIP_ASSIGN(ip, nlink, 2); 1448 if (DOINGSOFTDEP(tvp)) 1449 softdep_change_linkcnt(ip); 1450 if (cnp->cn_flags & ISWHITEOUT) { 1451 ip->i_flags |= UF_OPAQUE; 1452 DIP_ASSIGN(ip, flags, ip->i_flags); 1453 } 1454 1455 /* 1456 * Bump link count in parent directory to reflect work done below. 1457 * Should be done before reference is created so cleanup is 1458 * possible if we crash. 1459 */ 1460 dp->i_ffs_effnlink++; 1461 dp->i_nlink++; 1462 DIP_ASSIGN(dp, nlink, dp->i_nlink); 1463 dp->i_flag |= IN_CHANGE; 1464 if (DOINGSOFTDEP(dvp)) 1465 softdep_change_linkcnt(dp); 1466 if ((error = UFS_UPDATE(dvp, NULL, NULL, UPDATE_DIROP)) != 0) 1467 goto bad; 1468 1469 /* 1470 * Initialize directory with "." and ".." from static template. 1471 */ 1472 dirtemplate = mastertemplate; 1473 dirtemplate.dotdot_reclen = dirblksiz - dirtemplate.dot_reclen; 1474 dirtemplate.dot_ino = ufs_rw32(ip->i_number, UFS_MPNEEDSWAP(ump)); 1475 dirtemplate.dotdot_ino = ufs_rw32(dp->i_number, UFS_MPNEEDSWAP(ump)); 1476 dirtemplate.dot_reclen = ufs_rw16(dirtemplate.dot_reclen, 1477 UFS_MPNEEDSWAP(ump)); 1478 dirtemplate.dotdot_reclen = ufs_rw16(dirtemplate.dotdot_reclen, 1479 UFS_MPNEEDSWAP(ump)); 1480 if (ump->um_maxsymlinklen <= 0) { 1481 #if BYTE_ORDER == LITTLE_ENDIAN 1482 if (UFS_MPNEEDSWAP(ump) == 0) 1483 #else 1484 if (UFS_MPNEEDSWAP(ump) != 0) 1485 #endif 1486 { 1487 dirtemplate.dot_type = dirtemplate.dot_namlen; 1488 dirtemplate.dotdot_type = dirtemplate.dotdot_namlen; 1489 dirtemplate.dot_namlen = dirtemplate.dotdot_namlen = 0; 1490 } else 1491 dirtemplate.dot_type = dirtemplate.dotdot_type = 0; 1492 } 1493 if ((error = UFS_BALLOC(tvp, (off_t)0, dirblksiz, cnp->cn_cred, 1494 B_CLRBUF, &bp)) != 0) 1495 goto bad; 1496 ip->i_size = dirblksiz; 1497 DIP_ASSIGN(ip, size, dirblksiz); 1498 ip->i_flag |= IN_CHANGE | IN_UPDATE; 1499 uvm_vnp_setsize(tvp, ip->i_size); 1500 memcpy((void *)bp->b_data, (void *)&dirtemplate, sizeof dirtemplate); 1501 if (DOINGSOFTDEP(tvp)) { 1502 /* 1503 * Ensure that the entire newly allocated block is a 1504 * valid directory so that future growth within the 1505 * block does not have to ensure that the block is 1506 * written before the inode. 1507 */ 1508 blkoff = dirblksiz; 1509 while (blkoff < bp->b_bcount) { 1510 ((struct direct *) 1511 ((char *)bp->b_data + blkoff))->d_reclen = dirblksiz; 1512 blkoff += dirblksiz; 1513 } 1514 } 1515 /* 1516 * Directory set up, now install it's entry in the parent directory. 1517 * 1518 * If we are not doing soft dependencies, then we must write out the 1519 * buffer containing the new directory body before entering the new 1520 * name in the parent. If we are doing soft dependencies, then the 1521 * buffer containing the new directory body will be passed to and 1522 * released in the soft dependency code after the code has attached 1523 * an appropriate ordering dependency to the buffer which ensures that 1524 * the buffer is written before the new name is written in the parent. 1525 */ 1526 if (!DOINGSOFTDEP(tvp) && ((error = VOP_BWRITE(bp)) != 0)) 1527 goto bad; 1528 if ((error = UFS_UPDATE(tvp, NULL, NULL, UPDATE_DIROP)) != 0) { 1529 if (DOINGSOFTDEP(tvp)) 1530 (void)VOP_BWRITE(bp); 1531 goto bad; 1532 } 1533 newdir = pool_cache_get(ufs_direct_cache, PR_WAITOK); 1534 ufs_makedirentry(ip, cnp, newdir); 1535 error = ufs_direnter(dvp, tvp, newdir, cnp, bp); 1536 pool_cache_put(ufs_direct_cache, newdir); 1537 bad: 1538 if (error == 0) { 1539 VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK); 1540 UFS_WAPBL_END(dvp->v_mount); 1541 } else { 1542 dp->i_ffs_effnlink--; 1543 dp->i_nlink--; 1544 DIP_ASSIGN(dp, nlink, dp->i_nlink); 1545 dp->i_flag |= IN_CHANGE; 1546 UFS_WAPBL_UPDATE(dvp, NULL, NULL, UPDATE_DIROP); 1547 if (DOINGSOFTDEP(dvp)) 1548 softdep_change_linkcnt(dp); 1549 /* 1550 * No need to do an explicit UFS_TRUNCATE here, vrele will 1551 * do this for us because we set the link count to 0. 1552 */ 1553 ip->i_ffs_effnlink = 0; 1554 ip->i_nlink = 0; 1555 DIP_ASSIGN(ip, nlink, 0); 1556 ip->i_flag |= IN_CHANGE; 1557 #ifdef LFS 1558 /* If IN_ADIROP, account for it */ 1559 lfs_unmark_vnode(tvp); 1560 #endif 1561 UFS_WAPBL_UPDATE(tvp, NULL, NULL, UPDATE_DIROP); 1562 if (DOINGSOFTDEP(tvp)) 1563 softdep_change_linkcnt(ip); 1564 UFS_WAPBL_END(dvp->v_mount); 1565 vput(tvp); 1566 } 1567 out: 1568 PNBUF_PUT(cnp->cn_pnbuf); 1569 fstrans_done(dvp->v_mount); 1570 vput(dvp); 1571 return (error); 1572 } 1573 1574 int 1575 ufs_rmdir(void *v) 1576 { 1577 struct vop_rmdir_args /* { 1578 struct vnode *a_dvp; 1579 struct vnode *a_vp; 1580 struct componentname *a_cnp; 1581 } */ *ap = v; 1582 struct vnode *vp, *dvp; 1583 struct componentname *cnp; 1584 struct inode *ip, *dp; 1585 int error; 1586 bool pace; 1587 1588 vp = ap->a_vp; 1589 dvp = ap->a_dvp; 1590 cnp = ap->a_cnp; 1591 ip = VTOI(vp); 1592 dp = VTOI(dvp); 1593 /* 1594 * No rmdir "." or of mounted directories please. 1595 */ 1596 if (dp == ip || vp->v_mountedhere != NULL) { 1597 if (dp == ip) 1598 vrele(vp); 1599 else 1600 vput(vp); 1601 vput(vp); 1602 return (EINVAL); 1603 } 1604 1605 fstrans_start(dvp->v_mount, FSTRANS_SHARED); 1606 1607 /* 1608 * Do not remove a directory that is in the process of being renamed. 1609 * Verify that the directory is empty (and valid). (Rmdir ".." won't 1610 * be valid since ".." will contain a reference to the current 1611 * directory and thus be non-empty.) 1612 */ 1613 error = 0; 1614 if (ip->i_flag & IN_RENAME) { 1615 error = EINVAL; 1616 goto out; 1617 } 1618 if (ip->i_ffs_effnlink != 2 || 1619 !ufs_dirempty(ip, dp->i_number, cnp->cn_cred)) { 1620 error = ENOTEMPTY; 1621 goto out; 1622 } 1623 if ((dp->i_flags & APPEND) || 1624 (ip->i_flags & (IMMUTABLE | APPEND))) { 1625 error = EPERM; 1626 goto out; 1627 } 1628 error = UFS_WAPBL_BEGIN(dvp->v_mount); 1629 if (error) 1630 goto out; 1631 /* 1632 * Delete reference to directory before purging 1633 * inode. If we crash in between, the directory 1634 * will be reattached to lost+found, 1635 */ 1636 if (DOINGSOFTDEP(vp)) { 1637 dp->i_ffs_effnlink--; 1638 ip->i_ffs_effnlink--; 1639 softdep_change_linkcnt(dp); 1640 softdep_change_linkcnt(ip); 1641 } 1642 error = ufs_dirremove(dvp, ip, cnp->cn_flags, 1); 1643 if (error) { 1644 if (DOINGSOFTDEP(vp)) { 1645 dp->i_ffs_effnlink++; 1646 ip->i_ffs_effnlink++; 1647 softdep_change_linkcnt(dp); 1648 softdep_change_linkcnt(ip); 1649 } 1650 UFS_WAPBL_END(dvp->v_mount); 1651 goto out; 1652 } 1653 VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK); 1654 cache_purge(dvp); 1655 /* 1656 * Truncate inode. The only stuff left in the directory is "." and 1657 * "..". The "." reference is inconsequential since we're quashing 1658 * it. The soft dependency code will arrange to do these operations 1659 * after the parent directory entry has been deleted on disk, so 1660 * when running with that code we avoid doing them now. 1661 */ 1662 if (!DOINGSOFTDEP(vp)) { 1663 dp->i_nlink--; 1664 dp->i_ffs_effnlink--; 1665 DIP_ASSIGN(dp, nlink, dp->i_nlink); 1666 dp->i_flag |= IN_CHANGE; 1667 UFS_WAPBL_UPDATE(dvp, NULL, NULL, UPDATE_DIROP); 1668 ip->i_nlink--; 1669 ip->i_ffs_effnlink--; 1670 DIP_ASSIGN(ip, nlink, ip->i_nlink); 1671 ip->i_flag |= IN_CHANGE; 1672 error = UFS_TRUNCATE(vp, (off_t)0, IO_SYNC, cnp->cn_cred); 1673 } 1674 cache_purge(vp); 1675 /* 1676 * Unlock the log while we still have reference to unlinked 1677 * directory vp so that it will not get locked for recycling 1678 */ 1679 UFS_WAPBL_END(dvp->v_mount); 1680 #ifdef UFS_DIRHASH 1681 if (ip->i_dirhash != NULL) 1682 ufsdirhash_free(ip); 1683 #endif 1684 out: 1685 VN_KNOTE(vp, NOTE_DELETE); 1686 fstrans_done(dvp->v_mount); 1687 pace = DOINGSOFTDEP(dvp); 1688 vput(dvp); 1689 vput(vp); 1690 if (pace) { 1691 /* 1692 * Give the syncer some breathing room so that it 1693 * can flush removes. XXX 1694 */ 1695 softdep_pace_dirrem(); 1696 } 1697 return (error); 1698 } 1699 1700 /* 1701 * symlink -- make a symbolic link 1702 */ 1703 int 1704 ufs_symlink(void *v) 1705 { 1706 struct vop_symlink_args /* { 1707 struct vnode *a_dvp; 1708 struct vnode **a_vpp; 1709 struct componentname *a_cnp; 1710 struct vattr *a_vap; 1711 char *a_target; 1712 } */ *ap = v; 1713 struct vnode *vp, **vpp; 1714 struct inode *ip; 1715 int len, error; 1716 1717 vpp = ap->a_vpp; 1718 /* 1719 * UFS_WAPBL_BEGIN1(dvp->v_mount, dvp) performed by successful 1720 * ufs_makeinode 1721 */ 1722 fstrans_start(ap->a_dvp->v_mount, FSTRANS_SHARED); 1723 error = ufs_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp, 1724 vpp, ap->a_cnp); 1725 if (error) 1726 goto out; 1727 VN_KNOTE(ap->a_dvp, NOTE_WRITE); 1728 vp = *vpp; 1729 len = strlen(ap->a_target); 1730 ip = VTOI(vp); 1731 if (len < ip->i_ump->um_maxsymlinklen) { 1732 memcpy((char *)SHORTLINK(ip), ap->a_target, len); 1733 ip->i_size = len; 1734 DIP_ASSIGN(ip, size, len); 1735 uvm_vnp_setsize(vp, ip->i_size); 1736 ip->i_flag |= IN_CHANGE | IN_UPDATE; 1737 UFS_WAPBL_UPDATE(vp, NULL, NULL, 0); 1738 } else 1739 error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0, 1740 UIO_SYSSPACE, IO_NODELOCKED | IO_JOURNALLOCKED, 1741 ap->a_cnp->cn_cred, NULL, NULL); 1742 UFS_WAPBL_END1(ap->a_dvp->v_mount, ap->a_dvp); 1743 if (error) 1744 vput(vp); 1745 out: 1746 fstrans_done(ap->a_dvp->v_mount); 1747 return (error); 1748 } 1749 1750 /* 1751 * Vnode op for reading directories. 1752 * 1753 * This routine handles converting from the on-disk directory format 1754 * "struct direct" to the in-memory format "struct dirent" as well as 1755 * byte swapping the entries if necessary. 1756 */ 1757 int 1758 ufs_readdir(void *v) 1759 { 1760 struct vop_readdir_args /* { 1761 struct vnode *a_vp; 1762 struct uio *a_uio; 1763 kauth_cred_t a_cred; 1764 int *a_eofflag; 1765 off_t **a_cookies; 1766 int *ncookies; 1767 } */ *ap = v; 1768 struct vnode *vp = ap->a_vp; 1769 struct direct *cdp, *ecdp; 1770 struct dirent *ndp; 1771 char *cdbuf, *ndbuf, *endp; 1772 struct uio auio, *uio; 1773 struct iovec aiov; 1774 int error; 1775 size_t count, ccount, rcount; 1776 off_t off, *ccp; 1777 off_t startoff; 1778 size_t skipbytes; 1779 struct ufsmount *ump = VFSTOUFS(vp->v_mount); 1780 int nswap = UFS_MPNEEDSWAP(ump); 1781 #if BYTE_ORDER == LITTLE_ENDIAN 1782 int needswap = ump->um_maxsymlinklen <= 0 && nswap == 0; 1783 #else 1784 int needswap = ump->um_maxsymlinklen <= 0 && nswap != 0; 1785 #endif 1786 uio = ap->a_uio; 1787 count = uio->uio_resid; 1788 rcount = count - ((uio->uio_offset + count) & (ump->um_dirblksiz - 1)); 1789 1790 if (rcount < _DIRENT_MINSIZE(cdp) || count < _DIRENT_MINSIZE(ndp)) 1791 return EINVAL; 1792 1793 startoff = uio->uio_offset & ~(ump->um_dirblksiz - 1); 1794 skipbytes = uio->uio_offset - startoff; 1795 rcount += skipbytes; 1796 1797 auio.uio_iov = &aiov; 1798 auio.uio_iovcnt = 1; 1799 auio.uio_offset = startoff; 1800 auio.uio_resid = rcount; 1801 UIO_SETUP_SYSSPACE(&auio); 1802 auio.uio_rw = UIO_READ; 1803 cdbuf = malloc(rcount, M_TEMP, M_WAITOK); 1804 aiov.iov_base = cdbuf; 1805 aiov.iov_len = rcount; 1806 error = VOP_READ(vp, &auio, 0, ap->a_cred); 1807 if (error != 0) { 1808 free(cdbuf, M_TEMP); 1809 return error; 1810 } 1811 1812 rcount -= auio.uio_resid; 1813 1814 cdp = (struct direct *)(void *)cdbuf; 1815 ecdp = (struct direct *)(void *)&cdbuf[rcount]; 1816 1817 ndbuf = malloc(count, M_TEMP, M_WAITOK); 1818 ndp = (struct dirent *)(void *)ndbuf; 1819 endp = &ndbuf[count]; 1820 1821 off = uio->uio_offset; 1822 if (ap->a_cookies) { 1823 ccount = rcount / _DIRENT_RECLEN(cdp, 1); 1824 ccp = *(ap->a_cookies) = malloc(ccount * sizeof(*ccp), 1825 M_TEMP, M_WAITOK); 1826 } else { 1827 /* XXX: GCC */ 1828 ccount = 0; 1829 ccp = NULL; 1830 } 1831 1832 while (cdp < ecdp) { 1833 cdp->d_reclen = ufs_rw16(cdp->d_reclen, nswap); 1834 if (skipbytes > 0) { 1835 if (cdp->d_reclen <= skipbytes) { 1836 skipbytes -= cdp->d_reclen; 1837 cdp = _DIRENT_NEXT(cdp); 1838 continue; 1839 } 1840 /* 1841 * invalid cookie. 1842 */ 1843 error = EINVAL; 1844 goto out; 1845 } 1846 if (cdp->d_reclen == 0) { 1847 struct dirent *ondp = ndp; 1848 ndp->d_reclen = _DIRENT_MINSIZE(ndp); 1849 ndp = _DIRENT_NEXT(ndp); 1850 ondp->d_reclen = 0; 1851 cdp = ecdp; 1852 break; 1853 } 1854 if (needswap) { 1855 ndp->d_type = cdp->d_namlen; 1856 ndp->d_namlen = cdp->d_type; 1857 } else { 1858 ndp->d_type = cdp->d_type; 1859 ndp->d_namlen = cdp->d_namlen; 1860 } 1861 ndp->d_reclen = _DIRENT_RECLEN(ndp, ndp->d_namlen); 1862 if ((char *)(void *)ndp + ndp->d_reclen + 1863 _DIRENT_MINSIZE(ndp) > endp) 1864 break; 1865 ndp->d_fileno = ufs_rw32(cdp->d_ino, nswap); 1866 (void)memcpy(ndp->d_name, cdp->d_name, ndp->d_namlen); 1867 memset(&ndp->d_name[ndp->d_namlen], 0, 1868 ndp->d_reclen - _DIRENT_NAMEOFF(ndp) - ndp->d_namlen); 1869 off += cdp->d_reclen; 1870 if (ap->a_cookies) { 1871 KASSERT(ccp - *(ap->a_cookies) < ccount); 1872 *(ccp++) = off; 1873 } 1874 ndp = _DIRENT_NEXT(ndp); 1875 cdp = _DIRENT_NEXT(cdp); 1876 } 1877 1878 count = ((char *)(void *)ndp - ndbuf); 1879 error = uiomove(ndbuf, count, uio); 1880 out: 1881 if (ap->a_cookies) { 1882 if (error) { 1883 free(*(ap->a_cookies), M_TEMP); 1884 *(ap->a_cookies) = NULL; 1885 *(ap->a_ncookies) = 0; 1886 } else { 1887 *ap->a_ncookies = ccp - *(ap->a_cookies); 1888 } 1889 } 1890 uio->uio_offset = off; 1891 free(ndbuf, M_TEMP); 1892 free(cdbuf, M_TEMP); 1893 *ap->a_eofflag = VTOI(vp)->i_size <= uio->uio_offset; 1894 return error; 1895 } 1896 1897 /* 1898 * Return target name of a symbolic link 1899 */ 1900 int 1901 ufs_readlink(void *v) 1902 { 1903 struct vop_readlink_args /* { 1904 struct vnode *a_vp; 1905 struct uio *a_uio; 1906 kauth_cred_t a_cred; 1907 } */ *ap = v; 1908 struct vnode *vp = ap->a_vp; 1909 struct inode *ip = VTOI(vp); 1910 struct ufsmount *ump = VFSTOUFS(vp->v_mount); 1911 int isize; 1912 1913 isize = ip->i_size; 1914 if (isize < ump->um_maxsymlinklen || 1915 (ump->um_maxsymlinklen == 0 && DIP(ip, blocks) == 0)) { 1916 uiomove((char *)SHORTLINK(ip), isize, ap->a_uio); 1917 return (0); 1918 } 1919 return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred)); 1920 } 1921 1922 /* 1923 * Calculate the logical to physical mapping if not done already, 1924 * then call the device strategy routine. 1925 */ 1926 int 1927 ufs_strategy(void *v) 1928 { 1929 struct vop_strategy_args /* { 1930 struct vnode *a_vp; 1931 struct buf *a_bp; 1932 } */ *ap = v; 1933 struct buf *bp; 1934 struct vnode *vp; 1935 struct inode *ip; 1936 int error; 1937 1938 bp = ap->a_bp; 1939 vp = ap->a_vp; 1940 ip = VTOI(vp); 1941 if (vp->v_type == VBLK || vp->v_type == VCHR) 1942 panic("ufs_strategy: spec"); 1943 KASSERT(bp->b_bcount != 0); 1944 if (bp->b_blkno == bp->b_lblkno) { 1945 error = VOP_BMAP(vp, bp->b_lblkno, NULL, &bp->b_blkno, 1946 NULL); 1947 if (error) { 1948 bp->b_error = error; 1949 biodone(bp); 1950 return (error); 1951 } 1952 if (bp->b_blkno == -1) /* no valid data */ 1953 clrbuf(bp); 1954 } 1955 if (bp->b_blkno < 0) { /* block is not on disk */ 1956 biodone(bp); 1957 return (0); 1958 } 1959 vp = ip->i_devvp; 1960 return (VOP_STRATEGY(vp, bp)); 1961 } 1962 1963 /* 1964 * Print out the contents of an inode. 1965 */ 1966 int 1967 ufs_print(void *v) 1968 { 1969 struct vop_print_args /* { 1970 struct vnode *a_vp; 1971 } */ *ap = v; 1972 struct vnode *vp; 1973 struct inode *ip; 1974 1975 vp = ap->a_vp; 1976 ip = VTOI(vp); 1977 printf("tag VT_UFS, ino %llu, on dev %d, %d", 1978 (unsigned long long)ip->i_number, 1979 major(ip->i_dev), minor(ip->i_dev)); 1980 printf(" flags 0x%x, effnlink %d, nlink %d\n", 1981 ip->i_flag, ip->i_ffs_effnlink, ip->i_nlink); 1982 printf("\tmode 0%o, owner %d, group %d, size %qd", 1983 ip->i_mode, ip->i_uid, ip->i_gid, 1984 (long long)ip->i_size); 1985 if (vp->v_type == VFIFO) 1986 fifo_printinfo(vp); 1987 printf("\n"); 1988 return (0); 1989 } 1990 1991 /* 1992 * Read wrapper for special devices. 1993 */ 1994 int 1995 ufsspec_read(void *v) 1996 { 1997 struct vop_read_args /* { 1998 struct vnode *a_vp; 1999 struct uio *a_uio; 2000 int a_ioflag; 2001 kauth_cred_t a_cred; 2002 } */ *ap = v; 2003 2004 /* 2005 * Set access flag. 2006 */ 2007 if ((ap->a_vp->v_mount->mnt_flag & MNT_NODEVMTIME) == 0) 2008 VTOI(ap->a_vp)->i_flag |= IN_ACCESS; 2009 return (VOCALL (spec_vnodeop_p, VOFFSET(vop_read), ap)); 2010 } 2011 2012 /* 2013 * Write wrapper for special devices. 2014 */ 2015 int 2016 ufsspec_write(void *v) 2017 { 2018 struct vop_write_args /* { 2019 struct vnode *a_vp; 2020 struct uio *a_uio; 2021 int a_ioflag; 2022 kauth_cred_t a_cred; 2023 } */ *ap = v; 2024 2025 /* 2026 * Set update and change flags. 2027 */ 2028 if ((ap->a_vp->v_mount->mnt_flag & MNT_NODEVMTIME) == 0) 2029 VTOI(ap->a_vp)->i_flag |= IN_MODIFY; 2030 return (VOCALL (spec_vnodeop_p, VOFFSET(vop_write), ap)); 2031 } 2032 2033 /* 2034 * Close wrapper for special devices. 2035 * 2036 * Update the times on the inode then do device close. 2037 */ 2038 int 2039 ufsspec_close(void *v) 2040 { 2041 struct vop_close_args /* { 2042 struct vnode *a_vp; 2043 int a_fflag; 2044 kauth_cred_t a_cred; 2045 } */ *ap = v; 2046 struct vnode *vp; 2047 struct inode *ip; 2048 2049 vp = ap->a_vp; 2050 ip = VTOI(vp); 2051 if (vp->v_usecount > 1) 2052 UFS_ITIMES(vp, NULL, NULL, NULL); 2053 return (VOCALL (spec_vnodeop_p, VOFFSET(vop_close), ap)); 2054 } 2055 2056 /* 2057 * Read wrapper for fifo's 2058 */ 2059 int 2060 ufsfifo_read(void *v) 2061 { 2062 struct vop_read_args /* { 2063 struct vnode *a_vp; 2064 struct uio *a_uio; 2065 int a_ioflag; 2066 kauth_cred_t a_cred; 2067 } */ *ap = v; 2068 2069 /* 2070 * Set access flag. 2071 */ 2072 VTOI(ap->a_vp)->i_flag |= IN_ACCESS; 2073 return (VOCALL (fifo_vnodeop_p, VOFFSET(vop_read), ap)); 2074 } 2075 2076 /* 2077 * Write wrapper for fifo's. 2078 */ 2079 int 2080 ufsfifo_write(void *v) 2081 { 2082 struct vop_write_args /* { 2083 struct vnode *a_vp; 2084 struct uio *a_uio; 2085 int a_ioflag; 2086 kauth_cred_t a_cred; 2087 } */ *ap = v; 2088 2089 /* 2090 * Set update and change flags. 2091 */ 2092 VTOI(ap->a_vp)->i_flag |= IN_MODIFY; 2093 return (VOCALL (fifo_vnodeop_p, VOFFSET(vop_write), ap)); 2094 } 2095 2096 /* 2097 * Close wrapper for fifo's. 2098 * 2099 * Update the times on the inode then do device close. 2100 */ 2101 int 2102 ufsfifo_close(void *v) 2103 { 2104 struct vop_close_args /* { 2105 struct vnode *a_vp; 2106 int a_fflag; 2107 kauth_cred_t a_cred; 2108 } */ *ap = v; 2109 struct vnode *vp; 2110 struct inode *ip; 2111 2112 vp = ap->a_vp; 2113 ip = VTOI(vp); 2114 if (ap->a_vp->v_usecount > 1) 2115 UFS_ITIMES(vp, NULL, NULL, NULL); 2116 return (VOCALL (fifo_vnodeop_p, VOFFSET(vop_close), ap)); 2117 } 2118 2119 /* 2120 * Return POSIX pathconf information applicable to ufs filesystems. 2121 */ 2122 int 2123 ufs_pathconf(void *v) 2124 { 2125 struct vop_pathconf_args /* { 2126 struct vnode *a_vp; 2127 int a_name; 2128 register_t *a_retval; 2129 } */ *ap = v; 2130 2131 switch (ap->a_name) { 2132 case _PC_LINK_MAX: 2133 *ap->a_retval = LINK_MAX; 2134 return (0); 2135 case _PC_NAME_MAX: 2136 *ap->a_retval = NAME_MAX; 2137 return (0); 2138 case _PC_PATH_MAX: 2139 *ap->a_retval = PATH_MAX; 2140 return (0); 2141 case _PC_PIPE_BUF: 2142 *ap->a_retval = PIPE_BUF; 2143 return (0); 2144 case _PC_CHOWN_RESTRICTED: 2145 *ap->a_retval = 1; 2146 return (0); 2147 case _PC_NO_TRUNC: 2148 *ap->a_retval = 1; 2149 return (0); 2150 case _PC_SYNC_IO: 2151 *ap->a_retval = 1; 2152 return (0); 2153 case _PC_FILESIZEBITS: 2154 *ap->a_retval = 42; 2155 return (0); 2156 case _PC_SYMLINK_MAX: 2157 *ap->a_retval = MAXPATHLEN; 2158 return (0); 2159 case _PC_2_SYMLINKS: 2160 *ap->a_retval = 1; 2161 return (0); 2162 default: 2163 return (EINVAL); 2164 } 2165 /* NOTREACHED */ 2166 } 2167 2168 /* 2169 * Advisory record locking support 2170 */ 2171 int 2172 ufs_advlock(void *v) 2173 { 2174 struct vop_advlock_args /* { 2175 struct vnode *a_vp; 2176 void * a_id; 2177 int a_op; 2178 struct flock *a_fl; 2179 int a_flags; 2180 } */ *ap = v; 2181 struct inode *ip; 2182 2183 ip = VTOI(ap->a_vp); 2184 return lf_advlock(ap, &ip->i_lockf, ip->i_size); 2185 } 2186 2187 /* 2188 * Initialize the vnode associated with a new inode, handle aliased 2189 * vnodes. 2190 */ 2191 void 2192 ufs_vinit(struct mount *mntp, int (**specops)(void *), int (**fifoops)(void *), 2193 struct vnode **vpp) 2194 { 2195 struct timeval tv; 2196 struct inode *ip; 2197 struct vnode *vp; 2198 dev_t rdev; 2199 struct ufsmount *ump; 2200 2201 vp = *vpp; 2202 ip = VTOI(vp); 2203 switch(vp->v_type = IFTOVT(ip->i_mode)) { 2204 case VCHR: 2205 case VBLK: 2206 vp->v_op = specops; 2207 ump = ip->i_ump; 2208 if (ump->um_fstype == UFS1) 2209 rdev = (dev_t)ufs_rw32(ip->i_ffs1_rdev, 2210 UFS_MPNEEDSWAP(ump)); 2211 else 2212 rdev = (dev_t)ufs_rw64(ip->i_ffs2_rdev, 2213 UFS_MPNEEDSWAP(ump)); 2214 spec_node_init(vp, rdev); 2215 break; 2216 case VFIFO: 2217 vp->v_op = fifoops; 2218 break; 2219 case VNON: 2220 case VBAD: 2221 case VSOCK: 2222 case VLNK: 2223 case VDIR: 2224 case VREG: 2225 break; 2226 } 2227 if (ip->i_number == ROOTINO) 2228 vp->v_vflag |= VV_ROOT; 2229 /* 2230 * Initialize modrev times 2231 */ 2232 getmicrouptime(&tv); 2233 ip->i_modrev = (uint64_t)(uint)tv.tv_sec << 32 2234 | tv.tv_usec * 4294u; 2235 *vpp = vp; 2236 } 2237 2238 /* 2239 * Allocate a new inode. 2240 */ 2241 int 2242 ufs_makeinode(int mode, struct vnode *dvp, struct vnode **vpp, 2243 struct componentname *cnp) 2244 { 2245 struct inode *ip, *pdir; 2246 struct direct *newdir; 2247 struct vnode *tvp; 2248 int error, ismember = 0; 2249 2250 UFS_WAPBL_JUNLOCK_ASSERT(dvp->v_mount); 2251 2252 pdir = VTOI(dvp); 2253 #ifdef DIAGNOSTIC 2254 if ((cnp->cn_flags & HASBUF) == 0) 2255 panic("ufs_makeinode: no name"); 2256 #endif 2257 if ((mode & IFMT) == 0) 2258 mode |= IFREG; 2259 2260 if ((error = UFS_VALLOC(dvp, mode, cnp->cn_cred, vpp)) != 0) { 2261 PNBUF_PUT(cnp->cn_pnbuf); 2262 vput(dvp); 2263 return (error); 2264 } 2265 tvp = *vpp; 2266 ip = VTOI(tvp); 2267 ip->i_gid = pdir->i_gid; 2268 DIP_ASSIGN(ip, gid, ip->i_gid); 2269 ip->i_uid = kauth_cred_geteuid(cnp->cn_cred); 2270 DIP_ASSIGN(ip, uid, ip->i_uid); 2271 error = UFS_WAPBL_BEGIN1(dvp->v_mount, dvp); 2272 if (error) { 2273 /* 2274 * Note, we can't VOP_VFREE(tvp) here like we should 2275 * because we can't write to the disk. Instead, we leave 2276 * the vnode dangling from the journal. 2277 */ 2278 vput(tvp); 2279 PNBUF_PUT(cnp->cn_pnbuf); 2280 vput(dvp); 2281 return (error); 2282 } 2283 #ifdef QUOTA 2284 if ((error = chkiq(ip, 1, cnp->cn_cred, 0))) { 2285 UFS_VFREE(tvp, ip->i_number, mode); 2286 UFS_WAPBL_END1(dvp->v_mount, dvp); 2287 vput(tvp); 2288 PNBUF_PUT(cnp->cn_pnbuf); 2289 vput(dvp); 2290 return (error); 2291 } 2292 #endif 2293 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE; 2294 ip->i_mode = mode; 2295 DIP_ASSIGN(ip, mode, mode); 2296 tvp->v_type = IFTOVT(mode); /* Rest init'd in getnewvnode(). */ 2297 ip->i_ffs_effnlink = 1; 2298 ip->i_nlink = 1; 2299 DIP_ASSIGN(ip, nlink, 1); 2300 if (DOINGSOFTDEP(tvp)) 2301 softdep_change_linkcnt(ip); 2302 if ((ip->i_mode & ISGID) && (kauth_cred_ismember_gid(cnp->cn_cred, 2303 ip->i_gid, &ismember) != 0 || !ismember) && 2304 kauth_authorize_generic(cnp->cn_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2305 ip->i_mode &= ~ISGID; 2306 DIP_ASSIGN(ip, mode, ip->i_mode); 2307 } 2308 2309 if (cnp->cn_flags & ISWHITEOUT) { 2310 ip->i_flags |= UF_OPAQUE; 2311 DIP_ASSIGN(ip, flags, ip->i_flags); 2312 } 2313 2314 /* 2315 * Make sure inode goes to disk before directory entry. 2316 */ 2317 if ((error = UFS_UPDATE(tvp, NULL, NULL, UPDATE_DIROP)) != 0) 2318 goto bad; 2319 newdir = pool_cache_get(ufs_direct_cache, PR_WAITOK); 2320 ufs_makedirentry(ip, cnp, newdir); 2321 error = ufs_direnter(dvp, tvp, newdir, cnp, NULL); 2322 pool_cache_put(ufs_direct_cache, newdir); 2323 if (error) 2324 goto bad; 2325 if ((cnp->cn_flags & SAVESTART) == 0) 2326 PNBUF_PUT(cnp->cn_pnbuf); 2327 vput(dvp); 2328 *vpp = tvp; 2329 return (0); 2330 2331 bad: 2332 /* 2333 * Write error occurred trying to update the inode 2334 * or the directory so must deallocate the inode. 2335 */ 2336 ip->i_ffs_effnlink = 0; 2337 ip->i_nlink = 0; 2338 DIP_ASSIGN(ip, nlink, 0); 2339 ip->i_flag |= IN_CHANGE; 2340 #ifdef LFS 2341 /* If IN_ADIROP, account for it */ 2342 lfs_unmark_vnode(tvp); 2343 #endif 2344 UFS_WAPBL_UPDATE(tvp, NULL, NULL, 0); 2345 if (DOINGSOFTDEP(tvp)) 2346 softdep_change_linkcnt(ip); 2347 tvp->v_type = VNON; /* explodes later if VBLK */ 2348 UFS_WAPBL_END1(dvp->v_mount, dvp); 2349 vput(tvp); 2350 PNBUF_PUT(cnp->cn_pnbuf); 2351 vput(dvp); 2352 return (error); 2353 } 2354 2355 /* 2356 * Allocate len bytes at offset off. 2357 */ 2358 int 2359 ufs_gop_alloc(struct vnode *vp, off_t off, off_t len, int flags, 2360 kauth_cred_t cred) 2361 { 2362 struct inode *ip = VTOI(vp); 2363 int error, delta, bshift, bsize; 2364 UVMHIST_FUNC("ufs_gop_alloc"); UVMHIST_CALLED(ubchist); 2365 2366 error = 0; 2367 bshift = vp->v_mount->mnt_fs_bshift; 2368 bsize = 1 << bshift; 2369 2370 delta = off & (bsize - 1); 2371 off -= delta; 2372 len += delta; 2373 2374 while (len > 0) { 2375 bsize = MIN(bsize, len); 2376 2377 error = UFS_BALLOC(vp, off, bsize, cred, flags, NULL); 2378 if (error) { 2379 goto out; 2380 } 2381 2382 /* 2383 * increase file size now, UFS_BALLOC() requires that 2384 * EOF be up-to-date before each call. 2385 */ 2386 2387 if (ip->i_size < off + bsize) { 2388 UVMHIST_LOG(ubchist, "vp %p old 0x%x new 0x%x", 2389 vp, ip->i_size, off + bsize, 0); 2390 ip->i_size = off + bsize; 2391 DIP_ASSIGN(ip, size, ip->i_size); 2392 } 2393 2394 off += bsize; 2395 len -= bsize; 2396 } 2397 2398 out: 2399 UFS_WAPBL_UPDATE(vp, NULL, NULL, 0); 2400 return error; 2401 } 2402 2403 void 2404 ufs_gop_markupdate(struct vnode *vp, int flags) 2405 { 2406 u_int32_t mask = 0; 2407 2408 if ((flags & GOP_UPDATE_ACCESSED) != 0) { 2409 mask = IN_ACCESS; 2410 } 2411 if ((flags & GOP_UPDATE_MODIFIED) != 0) { 2412 if (vp->v_type == VREG) { 2413 mask |= IN_CHANGE | IN_UPDATE; 2414 } else { 2415 mask |= IN_MODIFY; 2416 } 2417 } 2418 if (mask) { 2419 struct inode *ip = VTOI(vp); 2420 2421 ip->i_flag |= mask; 2422 } 2423 } 2424