1 /* $NetBSD: lfs_alloc.c,v 1.50 2001/09/28 11:59:55 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 #if defined(_KERNEL_OPT) 74 #include "opt_quota.h" 75 #endif 76 77 #include <sys/param.h> 78 #include <sys/systm.h> 79 #include <sys/kernel.h> 80 #include <sys/buf.h> 81 #include <sys/vnode.h> 82 #include <sys/syslog.h> 83 #include <sys/mount.h> 84 #include <sys/malloc.h> 85 #include <sys/pool.h> 86 #include <sys/proc.h> 87 88 #include <ufs/ufs/quota.h> 89 #include <ufs/ufs/inode.h> 90 #include <ufs/ufs/ufsmount.h> 91 #include <ufs/ufs/ufs_extern.h> 92 93 #include <ufs/lfs/lfs.h> 94 #include <ufs/lfs/lfs_extern.h> 95 96 extern int lfs_dirvcount; 97 extern struct lock ufs_hashlock; 98 99 static int extend_ifile(struct lfs *, struct ucred *); 100 static int lfs_ialloc(struct lfs *, struct vnode *, ino_t, int, struct vnode **); 101 102 /* 103 * Allocate a particular inode with a particular version number, freeing 104 * any previous versions of this inode that may have gone before. 105 * Used by the roll-forward code. 106 * 107 * XXX this function does not have appropriate locking to be used on a live fs; 108 * XXX but something similar could probably be used for an "undelete" call. 109 */ 110 int 111 lfs_rf_valloc(struct lfs *fs, ino_t ino, int version, struct proc *p, 112 struct vnode **vpp) 113 { 114 IFILE *ifp; 115 struct buf *bp, *cbp; 116 struct vnode *vp; 117 struct inode *ip; 118 ino_t tino, oldnext; 119 int error; 120 CLEANERINFO *cip; 121 122 /* 123 * First, just try a vget. If the version number is the one we want, 124 * we don't have to do anything else. If the version number is wrong, 125 * take appropriate action. 126 */ 127 error = VFS_VGET(fs->lfs_ivnode->v_mount, ino, &vp); 128 if (error == 0) { 129 /* printf("lfs_rf_valloc[1]: ino %d vp %p\n", ino, vp); */ 130 131 *vpp = vp; 132 ip = VTOI(vp); 133 if (ip->i_ffs_gen == version) 134 return 0; 135 else if (ip->i_ffs_gen < version) { 136 VOP_TRUNCATE(vp, (off_t)0, 0, NOCRED, p); 137 ip->i_ffs_gen = version; 138 LFS_SET_UINO(ip, IN_CHANGE | IN_MODIFIED | IN_UPDATE); 139 return 0; 140 } else { 141 /* printf("ino %d: asked for version %d but got %d\n", 142 ino, version, ip->i_ffs_gen); */ 143 vput(vp); 144 *vpp = NULLVP; 145 return EEXIST; 146 } 147 } 148 149 /* 150 * The inode is not in use. Find it on the free list. 151 */ 152 /* If the Ifile is too short to contain this inum, extend it */ 153 while (VTOI(fs->lfs_ivnode)->i_ffs_size <= (ino / 154 fs->lfs_ifpb + fs->lfs_cleansz + fs->lfs_segtabsz) 155 << fs->lfs_bshift) { 156 extend_ifile(fs, NOCRED); 157 } 158 159 LFS_IENTRY(ifp, fs, ino, bp); 160 oldnext = ifp->if_nextfree; 161 ifp->if_version = version; 162 brelse(bp); 163 164 LFS_GET_HEADFREE(fs, cip, cbp, &ino); 165 if (ino) { 166 LFS_PUT_HEADFREE(fs, cip, cbp, oldnext); 167 } else { 168 tino = ino; 169 while(1) { 170 LFS_IENTRY(ifp, fs, tino, bp); 171 if (ifp->if_nextfree == ino || 172 ifp->if_nextfree == LFS_UNUSED_INUM) 173 break; 174 tino = ifp->if_nextfree; 175 brelse(bp); 176 } 177 if (ifp->if_nextfree == LFS_UNUSED_INUM) { 178 brelse(bp); 179 return ENOENT; 180 } 181 ifp->if_nextfree = oldnext; 182 VOP_BWRITE(bp); 183 } 184 185 error = lfs_ialloc(fs, fs->lfs_ivnode, ino, version, &vp); 186 if (error == 0) { 187 /* 188 * Make it VREG so we can put blocks on it. We will change 189 * this later if it turns out to be some other kind of file. 190 */ 191 ip = VTOI(vp); 192 ip->i_ffs_mode = IFREG; 193 ip->i_ffs_nlink = 1; 194 ip->i_ffs_effnlink = 1; 195 ufs_vinit(vp->v_mount, lfs_specop_p, lfs_fifoop_p, &vp); 196 ip = VTOI(vp); 197 198 /* printf("lfs_rf_valloc: ino %d vp %p\n", ino, vp); */ 199 200 /* The dirop-nature of this vnode is past */ 201 (void)lfs_vunref(vp); 202 --lfs_dirvcount; 203 vp->v_flag &= ~VDIROP; 204 --fs->lfs_nadirop; 205 ip->i_flag &= ~IN_ADIROP; 206 } 207 *vpp = vp; 208 return error; 209 } 210 211 static int 212 extend_ifile(struct lfs *fs, struct ucred *cred) 213 { 214 struct vnode *vp; 215 struct inode *ip; 216 IFILE *ifp; 217 IFILE_V1 *ifp_v1; 218 struct buf *bp, *cbp; 219 int error; 220 ufs_daddr_t i, blkno, max; 221 ino_t oldlast; 222 CLEANERINFO *cip; 223 224 vp = fs->lfs_ivnode; 225 (void)lfs_vref(vp); 226 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 227 ip = VTOI(vp); 228 blkno = lblkno(fs, ip->i_ffs_size); 229 if ((error = VOP_BALLOC(vp, ip->i_ffs_size, fs->lfs_bsize, cred, 0, 230 &bp)) != 0) { 231 VOP_UNLOCK(vp, 0); 232 lfs_vunref(vp); 233 return (error); 234 } 235 ip->i_ffs_size += fs->lfs_bsize; 236 uvm_vnp_setsize(vp, ip->i_ffs_size); 237 VOP_UNLOCK(vp, 0); 238 239 i = (blkno - fs->lfs_segtabsz - fs->lfs_cleansz) * 240 fs->lfs_ifpb; 241 LFS_GET_HEADFREE(fs, cip, cbp, &oldlast); 242 LFS_PUT_HEADFREE(fs, cip, cbp, i); 243 #ifdef DIAGNOSTIC 244 if(fs->lfs_free == LFS_UNUSED_INUM) 245 panic("inode 0 allocated [2]"); 246 #endif /* DIAGNOSTIC */ 247 max = i + fs->lfs_ifpb; 248 /* printf("extend ifile for ino %d--%d\n", i, max); */ 249 250 if(fs->lfs_version == 1) { 251 for (ifp_v1 = (IFILE_V1 *)bp->b_data; i < max; ++ifp_v1) { 252 ifp_v1->if_version = 1; 253 ifp_v1->if_daddr = LFS_UNUSED_DADDR; 254 ifp_v1->if_nextfree = ++i; 255 } 256 ifp_v1--; 257 ifp_v1->if_nextfree = oldlast; 258 } else { 259 for (ifp = (IFILE *)bp->b_data; i < max; ++ifp) { 260 ifp->if_version = 1; 261 ifp->if_daddr = LFS_UNUSED_DADDR; 262 ifp->if_nextfree = ++i; 263 } 264 ifp--; 265 ifp->if_nextfree = oldlast; 266 } 267 268 (void) VOP_BWRITE(bp); /* Ifile */ 269 lfs_vunref(vp); 270 271 return 0; 272 } 273 274 /* Allocate a new inode. */ 275 /* ARGSUSED */ 276 /* VOP_BWRITE 2i times */ 277 int 278 lfs_valloc(void *v) 279 { 280 struct vop_valloc_args /* { 281 struct vnode *a_pvp; 282 int a_mode; 283 struct ucred *a_cred; 284 struct vnode **a_vpp; 285 } */ *ap = v; 286 struct lfs *fs; 287 struct buf *bp, *cbp; 288 struct ifile *ifp; 289 ino_t new_ino; 290 int error; 291 int new_gen; 292 CLEANERINFO *cip; 293 294 fs = VTOI(ap->a_pvp)->i_lfs; 295 if (fs->lfs_ronly) 296 return EROFS; 297 *ap->a_vpp = NULL; 298 299 if (fs->lfs_version == 1) { 300 /* 301 * Use lfs_seglock here, instead of fs->lfs_freelock, to 302 * ensure that the free list is not changed in between 303 * the time that the ifile blocks are written to disk 304 * and the time that the superblock is written to disk. 305 */ 306 lfs_seglock(fs, SEGM_PROT); 307 } else { 308 lockmgr(&fs->lfs_freelock, LK_EXCLUSIVE, 0); 309 } 310 311 /* Get the head of the freelist. */ 312 LFS_GET_HEADFREE(fs, cip, cbp, &new_ino); 313 314 #ifdef DIAGNOSTIC 315 if(new_ino == LFS_UNUSED_INUM) { 316 #ifdef DEBUG 317 lfs_dump_super(fs); 318 #endif /* DEBUG */ 319 panic("inode 0 allocated [1]"); 320 } 321 #endif /* DIAGNOSTIC */ 322 #ifdef ALLOCPRINT 323 printf("lfs_valloc: allocate inode %d\n", new_ino); 324 #endif 325 326 /* 327 * Remove the inode from the free list and write the new start 328 * of the free list into the superblock. 329 */ 330 LFS_IENTRY(ifp, fs, new_ino, bp); 331 if (ifp->if_daddr != LFS_UNUSED_DADDR) 332 panic("lfs_valloc: inuse inode %d on the free list", new_ino); 333 LFS_PUT_HEADFREE(fs, cip, cbp, ifp->if_nextfree); 334 335 new_gen = ifp->if_version; /* version was updated by vfree */ 336 brelse(bp); 337 338 /* Extend IFILE so that the next lfs_valloc will succeed. */ 339 if (fs->lfs_free == LFS_UNUSED_INUM) { 340 if ((error = extend_ifile(fs, ap->a_cred)) != 0) { 341 LFS_PUT_HEADFREE(fs, cip, cbp, new_ino); 342 if (fs->lfs_version == 1) 343 lfs_segunlock(fs); 344 else 345 lockmgr(&fs->lfs_freelock, LK_RELEASE, 0); 346 return error; 347 } 348 } 349 #ifdef DIAGNOSTIC 350 if(fs->lfs_free == LFS_UNUSED_INUM) 351 panic("inode 0 allocated [3]"); 352 #endif /* DIAGNOSTIC */ 353 354 if (fs->lfs_version == 1) 355 lfs_segunlock(fs); 356 else 357 lockmgr(&fs->lfs_freelock, LK_RELEASE, 0); 358 359 return lfs_ialloc(fs, ap->a_pvp, new_ino, new_gen, ap->a_vpp); 360 } 361 362 static int 363 lfs_ialloc(struct lfs *fs, struct vnode *pvp, ino_t new_ino, int new_gen, 364 struct vnode **vpp) 365 { 366 struct inode *ip; 367 struct vnode *vp; 368 IFILE *ifp; 369 struct buf *bp, *cbp; 370 int error; 371 CLEANERINFO *cip; 372 373 error = getnewvnode(VT_LFS, pvp->v_mount, lfs_vnodeop_p, &vp); 374 /* printf("lfs_ialloc: ino %d vp %p error %d\n", new_ino, vp, error);*/ 375 if (error) 376 goto errout; 377 378 lockmgr(&ufs_hashlock, LK_EXCLUSIVE, 0); 379 /* Create an inode to associate with the vnode. */ 380 lfs_vcreate(pvp->v_mount, new_ino, vp); 381 382 ip = VTOI(vp); 383 LFS_SET_UINO(ip, IN_CHANGE | IN_MODIFIED); 384 /* Zero out the direct and indirect block addresses. */ 385 bzero(&ip->i_din, sizeof(ip->i_din)); 386 ip->i_din.ffs_din.di_inumber = new_ino; 387 388 /* Set a new generation number for this inode. */ 389 if (new_gen) 390 ip->i_ffs_gen = new_gen; 391 392 /* Insert into the inode hash table. */ 393 ufs_ihashins(ip); 394 lockmgr(&ufs_hashlock, LK_RELEASE, 0); 395 396 ufs_vinit(vp->v_mount, lfs_specop_p, lfs_fifoop_p, &vp); 397 ip = VTOI(vp); 398 /* printf("lfs_ialloc[2]: ino %d vp %p\n", new_ino, vp);*/ 399 400 *vpp = vp; 401 #if 1 402 if(!(vp->v_flag & VDIROP)) { 403 (void)lfs_vref(vp); 404 ++lfs_dirvcount; 405 } 406 vp->v_flag |= VDIROP; 407 408 if(!(ip->i_flag & IN_ADIROP)) 409 ++fs->lfs_nadirop; 410 ip->i_flag |= IN_ADIROP; 411 #endif 412 VREF(ip->i_devvp); 413 /* Set superblock modified bit and increment file count. */ 414 fs->lfs_fmod = 1; 415 ++fs->lfs_nfiles; 416 return (0); 417 418 errout: 419 /* 420 * Put the new inum back on the free list. 421 */ 422 LFS_IENTRY(ifp, fs, new_ino, bp); 423 ifp->if_daddr = LFS_UNUSED_DADDR; 424 LFS_GET_HEADFREE(fs, cip, cbp, &(ifp->if_nextfree)); 425 LFS_PUT_HEADFREE(fs, cip, cbp, new_ino); 426 (void) VOP_BWRITE(bp); /* Ifile */ 427 428 *vpp = NULLVP; 429 return (error); 430 } 431 432 /* Create a new vnode/inode pair and initialize what fields we can. */ 433 void 434 lfs_vcreate(struct mount *mp, ino_t ino, struct vnode *vp) 435 { 436 struct inode *ip; 437 struct ufsmount *ump; 438 #ifdef QUOTA 439 int i; 440 #endif 441 442 /* Get a pointer to the private mount structure. */ 443 ump = VFSTOUFS(mp); 444 445 /* Initialize the inode. */ 446 ip = pool_get(&lfs_inode_pool, PR_WAITOK); 447 vp->v_data = ip; 448 ip->i_vnode = vp; 449 ip->i_devvp = ump->um_devvp; 450 ip->i_dev = ump->um_dev; 451 ip->i_number = ip->i_din.ffs_din.di_inumber = ino; 452 ip->i_lfs = ump->um_lfs; 453 #ifdef QUOTA 454 for (i = 0; i < MAXQUOTAS; i++) 455 ip->i_dquot[i] = NODQUOT; 456 #endif 457 ip->i_lockf = 0; 458 ip->i_diroff = 0; 459 ip->i_ffs_mode = 0; 460 ip->i_ffs_size = 0; 461 ip->i_ffs_blocks = 0; 462 ip->i_lfs_effnblks = 0; 463 ip->i_flag = 0; 464 /* Why was IN_MODIFIED ever set here? */ 465 /* LFS_SET_UINO(ip, IN_CHANGE | IN_MODIFIED); */ 466 } 467 468 /* Free an inode. */ 469 /* ARGUSED */ 470 /* VOP_BWRITE 2i times */ 471 int 472 lfs_vfree(void *v) 473 { 474 struct vop_vfree_args /* { 475 struct vnode *a_pvp; 476 ino_t a_ino; 477 int a_mode; 478 } */ *ap = v; 479 SEGUSE *sup; 480 CLEANERINFO *cip; 481 struct buf *cbp, *bp; 482 struct ifile *ifp; 483 struct inode *ip; 484 struct vnode *vp; 485 struct lfs *fs; 486 ufs_daddr_t old_iaddr; 487 ino_t ino, otail; 488 extern int lfs_dirvcount; 489 490 /* Get the inode number and file system. */ 491 vp = ap->a_pvp; 492 ip = VTOI(vp); 493 fs = ip->i_lfs; 494 ino = ip->i_number; 495 496 /* Drain of pending writes */ 497 if (fs->lfs_version > 1 && WRITEINPROG(vp)) 498 tsleep(vp, (PRIBIO+1), "lfs_vfree", 0); 499 500 if (fs->lfs_version == 1) 501 lfs_seglock(fs, SEGM_PROT); 502 else 503 lockmgr(&fs->lfs_freelock, LK_EXCLUSIVE, 0); 504 505 if(vp->v_flag & VDIROP) { 506 --lfs_dirvcount; 507 vp->v_flag &= ~VDIROP; 508 wakeup(&lfs_dirvcount); 509 lfs_vunref(vp); 510 } 511 if (ip->i_flag & IN_ADIROP) { 512 --fs->lfs_nadirop; 513 ip->i_flag &= ~IN_ADIROP; 514 } 515 516 LFS_CLR_UINO(ip, IN_ACCESSED|IN_CLEANING|IN_MODIFIED); 517 ip->i_flag &= ~IN_ALLMOD; 518 519 /* 520 * Set the ifile's inode entry to unused, increment its version number 521 * and link it onto the free chain. 522 */ 523 LFS_IENTRY(ifp, fs, ino, bp); 524 old_iaddr = ifp->if_daddr; 525 ifp->if_daddr = LFS_UNUSED_DADDR; 526 ++ifp->if_version; 527 if (fs->lfs_version == 1) { 528 LFS_GET_HEADFREE(fs, cip, cbp, &(ifp->if_nextfree)); 529 LFS_PUT_HEADFREE(fs, cip, cbp, ino); 530 (void) VOP_BWRITE(bp); /* Ifile */ 531 } else { 532 ifp->if_nextfree = LFS_UNUSED_INUM; 533 /* 534 * XXX Writing the freed node here means that it might not 535 * XXX make it into the free list in the event of a crash 536 * XXX (the ifile could be written before the rest of this 537 * XXX completes). 538 */ 539 (void) VOP_BWRITE(bp); /* Ifile */ 540 LFS_GET_TAILFREE(fs, cip, cbp, &otail); 541 LFS_IENTRY(ifp, fs, otail, bp); 542 ifp->if_nextfree = ino; 543 VOP_BWRITE(bp); 544 LFS_PUT_TAILFREE(fs, cip, cbp, ino); 545 } 546 #ifdef DIAGNOSTIC 547 if(ino == LFS_UNUSED_INUM) { 548 panic("inode 0 freed"); 549 } 550 #endif /* DIAGNOSTIC */ 551 if (old_iaddr != LFS_UNUSED_DADDR) { 552 LFS_SEGENTRY(sup, fs, dtosn(fs, old_iaddr), bp); 553 #ifdef DIAGNOSTIC 554 if (sup->su_nbytes < DINODE_SIZE) { 555 printf("lfs_vfree: negative byte count" 556 " (segment %d short by %d)\n", 557 dtosn(fs, old_iaddr), 558 (int)DINODE_SIZE - sup->su_nbytes); 559 panic("lfs_vfree: negative byte count"); 560 sup->su_nbytes = DINODE_SIZE; 561 } 562 #endif 563 sup->su_nbytes -= DINODE_SIZE; 564 (void) VOP_BWRITE(bp); /* Ifile */ 565 } 566 567 /* Set superblock modified bit and decrement file count. */ 568 fs->lfs_fmod = 1; 569 --fs->lfs_nfiles; 570 571 if (fs->lfs_version == 1) 572 lfs_segunlock(fs); 573 else 574 lockmgr(&fs->lfs_freelock, LK_RELEASE, 0); 575 return (0); 576 } 577