1 /* $NetBSD: lfs_alloc.c,v 1.53 2001/11/23 21:44:26 chs Exp $ */ 2 3 /*- 4 * Copyright (c) 1999, 2000 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Konrad E. Schroder <perseant@hhhh.org>. 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 NetBSD 21 * Foundation, Inc. and its contributors. 22 * 4. Neither the name of The NetBSD Foundation nor the names of its 23 * contributors may be used to endorse or promote products derived 24 * from this software without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 27 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 28 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 29 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 30 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 31 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 32 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 33 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 34 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 35 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 36 * POSSIBILITY OF SUCH DAMAGE. 37 */ 38 /* 39 * Copyright (c) 1991, 1993 40 * The Regents of the University of California. All rights reserved. 41 * 42 * Redistribution and use in source and binary forms, with or without 43 * modification, are permitted provided that the following conditions 44 * are met: 45 * 1. Redistributions of source code must retain the above copyright 46 * notice, this list of conditions and the following disclaimer. 47 * 2. Redistributions in binary form must reproduce the above copyright 48 * notice, this list of conditions and the following disclaimer in the 49 * documentation and/or other materials provided with the distribution. 50 * 3. All advertising materials mentioning features or use of this software 51 * must display the following acknowledgement: 52 * This product includes software developed by the University of 53 * California, Berkeley and its contributors. 54 * 4. Neither the name of the University nor the names of its contributors 55 * may be used to endorse or promote products derived from this software 56 * without specific prior written permission. 57 * 58 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 59 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 60 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 61 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 62 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 63 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 64 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 65 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 66 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 67 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 68 * SUCH DAMAGE. 69 * 70 * @(#)lfs_alloc.c 8.4 (Berkeley) 1/4/94 71 */ 72 73 #include <sys/cdefs.h> 74 __KERNEL_RCSID(0, "$NetBSD: lfs_alloc.c,v 1.53 2001/11/23 21:44:26 chs Exp $"); 75 76 #if defined(_KERNEL_OPT) 77 #include "opt_quota.h" 78 #endif 79 80 #include <sys/param.h> 81 #include <sys/systm.h> 82 #include <sys/kernel.h> 83 #include <sys/buf.h> 84 #include <sys/vnode.h> 85 #include <sys/syslog.h> 86 #include <sys/mount.h> 87 #include <sys/malloc.h> 88 #include <sys/pool.h> 89 #include <sys/proc.h> 90 91 #include <ufs/ufs/quota.h> 92 #include <ufs/ufs/inode.h> 93 #include <ufs/ufs/ufsmount.h> 94 #include <ufs/ufs/ufs_extern.h> 95 96 #include <ufs/lfs/lfs.h> 97 #include <ufs/lfs/lfs_extern.h> 98 99 extern int lfs_dirvcount; 100 extern struct lock ufs_hashlock; 101 102 static int extend_ifile(struct lfs *, struct ucred *); 103 static int lfs_ialloc(struct lfs *, struct vnode *, ino_t, int, struct vnode **); 104 105 /* 106 * Allocate a particular inode with a particular version number, freeing 107 * any previous versions of this inode that may have gone before. 108 * Used by the roll-forward code. 109 * 110 * XXX this function does not have appropriate locking to be used on a live fs; 111 * XXX but something similar could probably be used for an "undelete" call. 112 */ 113 int 114 lfs_rf_valloc(struct lfs *fs, ino_t ino, int version, struct proc *p, 115 struct vnode **vpp) 116 { 117 IFILE *ifp; 118 struct buf *bp, *cbp; 119 struct vnode *vp; 120 struct inode *ip; 121 ino_t tino, oldnext; 122 int error; 123 CLEANERINFO *cip; 124 125 /* 126 * First, just try a vget. If the version number is the one we want, 127 * we don't have to do anything else. If the version number is wrong, 128 * take appropriate action. 129 */ 130 error = VFS_VGET(fs->lfs_ivnode->v_mount, ino, &vp); 131 if (error == 0) { 132 /* printf("lfs_rf_valloc[1]: ino %d vp %p\n", ino, vp); */ 133 134 *vpp = vp; 135 ip = VTOI(vp); 136 if (ip->i_ffs_gen == version) 137 return 0; 138 else if (ip->i_ffs_gen < version) { 139 VOP_TRUNCATE(vp, (off_t)0, 0, NOCRED, p); 140 ip->i_ffs_gen = version; 141 LFS_SET_UINO(ip, IN_CHANGE | IN_MODIFIED | IN_UPDATE); 142 return 0; 143 } else { 144 /* printf("ino %d: asked for version %d but got %d\n", 145 ino, version, ip->i_ffs_gen); */ 146 vput(vp); 147 *vpp = NULLVP; 148 return EEXIST; 149 } 150 } 151 152 /* 153 * The inode is not in use. Find it on the free list. 154 */ 155 /* If the Ifile is too short to contain this inum, extend it */ 156 while (VTOI(fs->lfs_ivnode)->i_ffs_size <= (ino / 157 fs->lfs_ifpb + fs->lfs_cleansz + fs->lfs_segtabsz) 158 << fs->lfs_bshift) { 159 extend_ifile(fs, NOCRED); 160 } 161 162 LFS_IENTRY(ifp, fs, ino, bp); 163 oldnext = ifp->if_nextfree; 164 ifp->if_version = version; 165 brelse(bp); 166 167 LFS_GET_HEADFREE(fs, cip, cbp, &ino); 168 if (ino) { 169 LFS_PUT_HEADFREE(fs, cip, cbp, oldnext); 170 } else { 171 tino = ino; 172 while (1) { 173 LFS_IENTRY(ifp, fs, tino, bp); 174 if (ifp->if_nextfree == ino || 175 ifp->if_nextfree == LFS_UNUSED_INUM) 176 break; 177 tino = ifp->if_nextfree; 178 brelse(bp); 179 } 180 if (ifp->if_nextfree == LFS_UNUSED_INUM) { 181 brelse(bp); 182 return ENOENT; 183 } 184 ifp->if_nextfree = oldnext; 185 VOP_BWRITE(bp); 186 } 187 188 error = lfs_ialloc(fs, fs->lfs_ivnode, ino, version, &vp); 189 if (error == 0) { 190 /* 191 * Make it VREG so we can put blocks on it. We will change 192 * this later if it turns out to be some other kind of file. 193 */ 194 ip = VTOI(vp); 195 ip->i_ffs_mode = IFREG; 196 ip->i_ffs_nlink = 1; 197 ip->i_ffs_effnlink = 1; 198 ufs_vinit(vp->v_mount, lfs_specop_p, lfs_fifoop_p, &vp); 199 ip = VTOI(vp); 200 201 /* printf("lfs_rf_valloc: ino %d vp %p\n", ino, vp); */ 202 203 /* The dirop-nature of this vnode is past */ 204 (void)lfs_vunref(vp); 205 --lfs_dirvcount; 206 vp->v_flag &= ~VDIROP; 207 --fs->lfs_nadirop; 208 ip->i_flag &= ~IN_ADIROP; 209 } 210 *vpp = vp; 211 return error; 212 } 213 214 static int 215 extend_ifile(struct lfs *fs, struct ucred *cred) 216 { 217 struct vnode *vp; 218 struct inode *ip; 219 IFILE *ifp; 220 IFILE_V1 *ifp_v1; 221 struct buf *bp, *cbp; 222 int error; 223 ufs_daddr_t i, blkno, max; 224 ino_t oldlast; 225 CLEANERINFO *cip; 226 227 vp = fs->lfs_ivnode; 228 (void)lfs_vref(vp); 229 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 230 ip = VTOI(vp); 231 blkno = lblkno(fs, ip->i_ffs_size); 232 if ((error = VOP_BALLOC(vp, ip->i_ffs_size, fs->lfs_bsize, cred, 0, 233 &bp)) != 0) { 234 VOP_UNLOCK(vp, 0); 235 lfs_vunref(vp); 236 return (error); 237 } 238 ip->i_ffs_size += fs->lfs_bsize; 239 uvm_vnp_setsize(vp, ip->i_ffs_size); 240 VOP_UNLOCK(vp, 0); 241 242 i = (blkno - fs->lfs_segtabsz - fs->lfs_cleansz) * 243 fs->lfs_ifpb; 244 LFS_GET_HEADFREE(fs, cip, cbp, &oldlast); 245 LFS_PUT_HEADFREE(fs, cip, cbp, i); 246 #ifdef DIAGNOSTIC 247 if (fs->lfs_free == LFS_UNUSED_INUM) 248 panic("inode 0 allocated [2]"); 249 #endif /* DIAGNOSTIC */ 250 max = i + fs->lfs_ifpb; 251 /* printf("extend ifile for ino %d--%d\n", i, max); */ 252 253 if (fs->lfs_version == 1) { 254 for (ifp_v1 = (IFILE_V1 *)bp->b_data; i < max; ++ifp_v1) { 255 ifp_v1->if_version = 1; 256 ifp_v1->if_daddr = LFS_UNUSED_DADDR; 257 ifp_v1->if_nextfree = ++i; 258 } 259 ifp_v1--; 260 ifp_v1->if_nextfree = oldlast; 261 } else { 262 for (ifp = (IFILE *)bp->b_data; i < max; ++ifp) { 263 ifp->if_version = 1; 264 ifp->if_daddr = LFS_UNUSED_DADDR; 265 ifp->if_nextfree = ++i; 266 } 267 ifp--; 268 ifp->if_nextfree = oldlast; 269 } 270 271 (void) VOP_BWRITE(bp); /* Ifile */ 272 lfs_vunref(vp); 273 274 return 0; 275 } 276 277 /* Allocate a new inode. */ 278 /* ARGSUSED */ 279 /* VOP_BWRITE 2i times */ 280 int 281 lfs_valloc(void *v) 282 { 283 struct vop_valloc_args /* { 284 struct vnode *a_pvp; 285 int a_mode; 286 struct ucred *a_cred; 287 struct vnode **a_vpp; 288 } */ *ap = v; 289 struct lfs *fs; 290 struct buf *bp, *cbp; 291 struct ifile *ifp; 292 ino_t new_ino; 293 int error; 294 int new_gen; 295 CLEANERINFO *cip; 296 297 fs = VTOI(ap->a_pvp)->i_lfs; 298 if (fs->lfs_ronly) 299 return EROFS; 300 *ap->a_vpp = NULL; 301 302 if (fs->lfs_version == 1) { 303 /* 304 * Use lfs_seglock here, instead of fs->lfs_freelock, to 305 * ensure that the free list is not changed in between 306 * the time that the ifile blocks are written to disk 307 * and the time that the superblock is written to disk. 308 */ 309 lfs_seglock(fs, SEGM_PROT); 310 } else { 311 lockmgr(&fs->lfs_freelock, LK_EXCLUSIVE, 0); 312 } 313 314 /* Get the head of the freelist. */ 315 LFS_GET_HEADFREE(fs, cip, cbp, &new_ino); 316 317 #ifdef DIAGNOSTIC 318 if (new_ino == LFS_UNUSED_INUM) { 319 #ifdef DEBUG 320 lfs_dump_super(fs); 321 #endif /* DEBUG */ 322 panic("inode 0 allocated [1]"); 323 } 324 #endif /* DIAGNOSTIC */ 325 #ifdef ALLOCPRINT 326 printf("lfs_valloc: allocate inode %d\n", new_ino); 327 #endif 328 329 /* 330 * Remove the inode from the free list and write the new start 331 * of the free list into the superblock. 332 */ 333 LFS_IENTRY(ifp, fs, new_ino, bp); 334 if (ifp->if_daddr != LFS_UNUSED_DADDR) 335 panic("lfs_valloc: inuse inode %d on the free list", new_ino); 336 LFS_PUT_HEADFREE(fs, cip, cbp, ifp->if_nextfree); 337 338 new_gen = ifp->if_version; /* version was updated by vfree */ 339 brelse(bp); 340 341 /* Extend IFILE so that the next lfs_valloc will succeed. */ 342 if (fs->lfs_free == LFS_UNUSED_INUM) { 343 if ((error = extend_ifile(fs, ap->a_cred)) != 0) { 344 LFS_PUT_HEADFREE(fs, cip, cbp, new_ino); 345 if (fs->lfs_version == 1) 346 lfs_segunlock(fs); 347 else 348 lockmgr(&fs->lfs_freelock, LK_RELEASE, 0); 349 return error; 350 } 351 } 352 #ifdef DIAGNOSTIC 353 if (fs->lfs_free == LFS_UNUSED_INUM) 354 panic("inode 0 allocated [3]"); 355 #endif /* DIAGNOSTIC */ 356 357 if (fs->lfs_version == 1) 358 lfs_segunlock(fs); 359 else 360 lockmgr(&fs->lfs_freelock, LK_RELEASE, 0); 361 362 return lfs_ialloc(fs, ap->a_pvp, new_ino, new_gen, ap->a_vpp); 363 } 364 365 static int 366 lfs_ialloc(struct lfs *fs, struct vnode *pvp, ino_t new_ino, int new_gen, 367 struct vnode **vpp) 368 { 369 struct inode *ip; 370 struct vnode *vp; 371 IFILE *ifp; 372 struct buf *bp, *cbp; 373 int error; 374 CLEANERINFO *cip; 375 376 error = getnewvnode(VT_LFS, pvp->v_mount, lfs_vnodeop_p, &vp); 377 /* printf("lfs_ialloc: ino %d vp %p error %d\n", new_ino, vp, error);*/ 378 if (error) 379 goto errout; 380 381 lockmgr(&ufs_hashlock, LK_EXCLUSIVE, 0); 382 /* Create an inode to associate with the vnode. */ 383 lfs_vcreate(pvp->v_mount, new_ino, vp); 384 385 ip = VTOI(vp); 386 LFS_SET_UINO(ip, IN_CHANGE | IN_MODIFIED); 387 /* Zero out the direct and indirect block addresses. */ 388 bzero(&ip->i_din, sizeof(ip->i_din)); 389 ip->i_din.ffs_din.di_inumber = new_ino; 390 391 /* Set a new generation number for this inode. */ 392 if (new_gen) 393 ip->i_ffs_gen = new_gen; 394 395 /* Insert into the inode hash table. */ 396 ufs_ihashins(ip); 397 lockmgr(&ufs_hashlock, LK_RELEASE, 0); 398 399 ufs_vinit(vp->v_mount, lfs_specop_p, lfs_fifoop_p, &vp); 400 ip = VTOI(vp); 401 /* printf("lfs_ialloc[2]: ino %d vp %p\n", new_ino, vp);*/ 402 403 uvm_vnp_setsize(vp, 0); 404 *vpp = vp; 405 #if 1 406 if (!(vp->v_flag & VDIROP)) { 407 (void)lfs_vref(vp); 408 ++lfs_dirvcount; 409 } 410 vp->v_flag |= VDIROP; 411 412 if (!(ip->i_flag & IN_ADIROP)) 413 ++fs->lfs_nadirop; 414 ip->i_flag |= IN_ADIROP; 415 #endif 416 VREF(ip->i_devvp); 417 /* Set superblock modified bit and increment file count. */ 418 fs->lfs_fmod = 1; 419 ++fs->lfs_nfiles; 420 return (0); 421 422 errout: 423 /* 424 * Put the new inum back on the free list. 425 */ 426 LFS_IENTRY(ifp, fs, new_ino, bp); 427 ifp->if_daddr = LFS_UNUSED_DADDR; 428 LFS_GET_HEADFREE(fs, cip, cbp, &(ifp->if_nextfree)); 429 LFS_PUT_HEADFREE(fs, cip, cbp, new_ino); 430 (void) VOP_BWRITE(bp); /* Ifile */ 431 432 *vpp = NULLVP; 433 return (error); 434 } 435 436 /* Create a new vnode/inode pair and initialize what fields we can. */ 437 void 438 lfs_vcreate(struct mount *mp, ino_t ino, struct vnode *vp) 439 { 440 struct inode *ip; 441 struct ufsmount *ump; 442 #ifdef QUOTA 443 int i; 444 #endif 445 446 /* Get a pointer to the private mount structure. */ 447 ump = VFSTOUFS(mp); 448 449 /* Initialize the inode. */ 450 ip = pool_get(&lfs_inode_pool, PR_WAITOK); 451 vp->v_data = ip; 452 ip->i_vnode = vp; 453 ip->i_devvp = ump->um_devvp; 454 ip->i_dev = ump->um_dev; 455 ip->i_number = ip->i_din.ffs_din.di_inumber = ino; 456 ip->i_lfs = ump->um_lfs; 457 #ifdef QUOTA 458 for (i = 0; i < MAXQUOTAS; i++) 459 ip->i_dquot[i] = NODQUOT; 460 #endif 461 ip->i_lockf = 0; 462 ip->i_diroff = 0; 463 ip->i_ffs_mode = 0; 464 ip->i_ffs_size = 0; 465 ip->i_ffs_blocks = 0; 466 ip->i_lfs_effnblks = 0; 467 ip->i_flag = 0; 468 /* Why was IN_MODIFIED ever set here? */ 469 /* LFS_SET_UINO(ip, IN_CHANGE | IN_MODIFIED); */ 470 } 471 472 /* Free an inode. */ 473 /* ARGUSED */ 474 /* VOP_BWRITE 2i times */ 475 int 476 lfs_vfree(void *v) 477 { 478 struct vop_vfree_args /* { 479 struct vnode *a_pvp; 480 ino_t a_ino; 481 int a_mode; 482 } */ *ap = v; 483 SEGUSE *sup; 484 CLEANERINFO *cip; 485 struct buf *cbp, *bp; 486 struct ifile *ifp; 487 struct inode *ip; 488 struct vnode *vp; 489 struct lfs *fs; 490 ufs_daddr_t old_iaddr; 491 ino_t ino, otail; 492 extern int lfs_dirvcount; 493 494 /* Get the inode number and file system. */ 495 vp = ap->a_pvp; 496 ip = VTOI(vp); 497 fs = ip->i_lfs; 498 ino = ip->i_number; 499 500 /* Drain of pending writes */ 501 if (fs->lfs_version > 1 && WRITEINPROG(vp)) 502 tsleep(vp, (PRIBIO+1), "lfs_vfree", 0); 503 504 if (fs->lfs_version == 1) 505 lfs_seglock(fs, SEGM_PROT); 506 else 507 lockmgr(&fs->lfs_freelock, LK_EXCLUSIVE, 0); 508 509 if (vp->v_flag & VDIROP) { 510 --lfs_dirvcount; 511 vp->v_flag &= ~VDIROP; 512 wakeup(&lfs_dirvcount); 513 lfs_vunref(vp); 514 } 515 if (ip->i_flag & IN_ADIROP) { 516 --fs->lfs_nadirop; 517 ip->i_flag &= ~IN_ADIROP; 518 } 519 520 LFS_CLR_UINO(ip, IN_ACCESSED|IN_CLEANING|IN_MODIFIED); 521 ip->i_flag &= ~IN_ALLMOD; 522 523 /* 524 * Set the ifile's inode entry to unused, increment its version number 525 * and link it onto the free chain. 526 */ 527 LFS_IENTRY(ifp, fs, ino, bp); 528 old_iaddr = ifp->if_daddr; 529 ifp->if_daddr = LFS_UNUSED_DADDR; 530 ++ifp->if_version; 531 if (fs->lfs_version == 1) { 532 LFS_GET_HEADFREE(fs, cip, cbp, &(ifp->if_nextfree)); 533 LFS_PUT_HEADFREE(fs, cip, cbp, ino); 534 (void) VOP_BWRITE(bp); /* Ifile */ 535 } else { 536 ifp->if_nextfree = LFS_UNUSED_INUM; 537 /* 538 * XXX Writing the freed node here means that it might not 539 * XXX make it into the free list in the event of a crash 540 * XXX (the ifile could be written before the rest of this 541 * XXX completes). 542 */ 543 (void) VOP_BWRITE(bp); /* Ifile */ 544 LFS_GET_TAILFREE(fs, cip, cbp, &otail); 545 LFS_IENTRY(ifp, fs, otail, bp); 546 ifp->if_nextfree = ino; 547 VOP_BWRITE(bp); 548 LFS_PUT_TAILFREE(fs, cip, cbp, ino); 549 } 550 #ifdef DIAGNOSTIC 551 if (ino == LFS_UNUSED_INUM) { 552 panic("inode 0 freed"); 553 } 554 #endif /* DIAGNOSTIC */ 555 if (old_iaddr != LFS_UNUSED_DADDR) { 556 LFS_SEGENTRY(sup, fs, dtosn(fs, old_iaddr), bp); 557 #ifdef DIAGNOSTIC 558 if (sup->su_nbytes < DINODE_SIZE) { 559 printf("lfs_vfree: negative byte count" 560 " (segment %d short by %d)\n", 561 dtosn(fs, old_iaddr), 562 (int)DINODE_SIZE - sup->su_nbytes); 563 panic("lfs_vfree: negative byte count"); 564 sup->su_nbytes = DINODE_SIZE; 565 } 566 #endif 567 sup->su_nbytes -= DINODE_SIZE; 568 (void) VOP_BWRITE(bp); /* Ifile */ 569 } 570 571 /* Set superblock modified bit and decrement file count. */ 572 fs->lfs_fmod = 1; 573 --fs->lfs_nfiles; 574 575 if (fs->lfs_version == 1) 576 lfs_segunlock(fs); 577 else 578 lockmgr(&fs->lfs_freelock, LK_RELEASE, 0); 579 return (0); 580 } 581