1 /* $NetBSD: lfs_alloc.c,v 1.48 2001/07/13 20:30:23 perseant 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 87 #include <ufs/ufs/quota.h> 88 #include <ufs/ufs/inode.h> 89 #include <ufs/ufs/ufsmount.h> 90 #include <ufs/ufs/ufs_extern.h> 91 92 #include <ufs/lfs/lfs.h> 93 #include <ufs/lfs/lfs_extern.h> 94 95 extern int lfs_dirvcount; 96 extern struct lock ufs_hashlock; 97 98 static int extend_ifile(struct lfs *, struct ucred *); 99 static int lfs_ialloc(struct lfs *, struct vnode *, ino_t, int, struct vnode **); 100 101 /* 102 * Allocate a particular inode with a particular version number, freeing 103 * any previous versions of this inode that may have gone before. 104 * Used by the roll-forward code. 105 * 106 * XXX this function does not have appropriate locking to be used on a live fs; 107 * XXX but something similar could probably be used for an "undelete" call. 108 */ 109 int 110 lfs_rf_valloc(struct lfs *fs, ino_t ino, int version, struct proc *p, 111 struct vnode **vpp) 112 { 113 IFILE *ifp; 114 struct buf *bp, *cbp; 115 struct vnode *vp; 116 struct inode *ip; 117 ino_t tino, oldnext; 118 int error; 119 CLEANERINFO *cip; 120 121 /* 122 * First, just try a vget. If the version number is the one we want, 123 * we don't have to do anything else. If the version number is wrong, 124 * take appropriate action. 125 */ 126 error = VFS_VGET(fs->lfs_ivnode->v_mount, ino, &vp); 127 if (error == 0) { 128 /* printf("lfs_rf_valloc[1]: ino %d vp %p\n", ino, vp); */ 129 130 *vpp = vp; 131 ip = VTOI(vp); 132 if (ip->i_ffs_gen == version) 133 return 0; 134 else if (ip->i_ffs_gen < version) { 135 VOP_TRUNCATE(vp, (off_t)0, 0, NOCRED, p); 136 ip->i_ffs_gen = version; 137 LFS_SET_UINO(ip, IN_CHANGE | IN_MODIFIED | IN_UPDATE); 138 return 0; 139 } else { 140 /* printf("ino %d: asked for version %d but got %d\n", 141 ino, version, ip->i_ffs_gen); */ 142 vput(vp); 143 *vpp = NULLVP; 144 return EEXIST; 145 } 146 } 147 148 /* 149 * The inode is not in use. Find it on the free list. 150 */ 151 /* If the Ifile is too short to contain this inum, extend it */ 152 while (VTOI(fs->lfs_ivnode)->i_ffs_size <= (ino / 153 fs->lfs_ifpb + fs->lfs_cleansz + fs->lfs_segtabsz) 154 << fs->lfs_bshift) { 155 extend_ifile(fs, NOCRED); 156 } 157 158 LFS_IENTRY(ifp, fs, ino, bp); 159 oldnext = ifp->if_nextfree; 160 ifp->if_version = version; 161 brelse(bp); 162 163 LFS_GET_HEADFREE(fs, cip, cbp, &ino); 164 if (ino) { 165 LFS_PUT_HEADFREE(fs, cip, cbp, oldnext); 166 } else { 167 tino = ino; 168 while(1) { 169 LFS_IENTRY(ifp, fs, tino, bp); 170 if (ifp->if_nextfree == ino || 171 ifp->if_nextfree == LFS_UNUSED_INUM) 172 break; 173 tino = ifp->if_nextfree; 174 brelse(bp); 175 } 176 if (ifp->if_nextfree == LFS_UNUSED_INUM) { 177 brelse(bp); 178 return ENOENT; 179 } 180 ifp->if_nextfree = oldnext; 181 VOP_BWRITE(bp); 182 } 183 184 error = lfs_ialloc(fs, fs->lfs_ivnode, ino, version, &vp); 185 if (error == 0) { 186 /* 187 * Make it VREG so we can put blocks on it. We will change 188 * this later if it turns out to be some other kind of file. 189 */ 190 ip = VTOI(vp); 191 ip->i_ffs_mode = IFREG; 192 ip->i_ffs_nlink = 1; 193 ip->i_ffs_effnlink = 1; 194 ufs_vinit(vp->v_mount, lfs_specop_p, lfs_fifoop_p, &vp); 195 ip = VTOI(vp); 196 197 /* printf("lfs_rf_valloc: ino %d vp %p\n", ino, vp); */ 198 199 /* The dirop-nature of this vnode is past */ 200 (void)lfs_vunref(vp); 201 --lfs_dirvcount; 202 vp->v_flag &= ~VDIROP; 203 --fs->lfs_nadirop; 204 ip->i_flag &= ~IN_ADIROP; 205 } 206 *vpp = vp; 207 return error; 208 } 209 210 static int 211 extend_ifile(struct lfs *fs, struct ucred *cred) 212 { 213 struct vnode *vp; 214 struct inode *ip; 215 IFILE *ifp; 216 IFILE_V1 *ifp_v1; 217 struct buf *bp, *cbp; 218 int error; 219 ufs_daddr_t i, blkno, max; 220 ino_t oldlast; 221 CLEANERINFO *cip; 222 223 vp = fs->lfs_ivnode; 224 (void)lfs_vref(vp); 225 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 226 ip = VTOI(vp); 227 blkno = lblkno(fs, ip->i_ffs_size); 228 if ((error = VOP_BALLOC(vp, ip->i_ffs_size, fs->lfs_bsize, cred, 0, 229 &bp)) != 0) { 230 VOP_UNLOCK(vp, 0); 231 lfs_vunref(vp); 232 return (error); 233 } 234 ip->i_ffs_size += fs->lfs_bsize; 235 uvm_vnp_setsize(vp, ip->i_ffs_size); 236 VOP_UNLOCK(vp, 0); 237 238 i = (blkno - fs->lfs_segtabsz - fs->lfs_cleansz) * 239 fs->lfs_ifpb; 240 LFS_GET_HEADFREE(fs, cip, cbp, &oldlast); 241 LFS_PUT_HEADFREE(fs, cip, cbp, i); 242 #ifdef DIAGNOSTIC 243 if(fs->lfs_free == LFS_UNUSED_INUM) 244 panic("inode 0 allocated [2]"); 245 #endif /* DIAGNOSTIC */ 246 max = i + fs->lfs_ifpb; 247 /* printf("extend ifile for ino %d--%d\n", i, max); */ 248 249 if(fs->lfs_version == 1) { 250 for (ifp_v1 = (IFILE_V1 *)bp->b_data; i < max; ++ifp_v1) { 251 ifp_v1->if_version = 1; 252 ifp_v1->if_daddr = LFS_UNUSED_DADDR; 253 ifp_v1->if_nextfree = ++i; 254 } 255 ifp_v1--; 256 ifp_v1->if_nextfree = oldlast; 257 } else { 258 for (ifp = (IFILE *)bp->b_data; i < max; ++ifp) { 259 ifp->if_version = 1; 260 ifp->if_daddr = LFS_UNUSED_DADDR; 261 ifp->if_nextfree = ++i; 262 } 263 ifp--; 264 ifp->if_nextfree = oldlast; 265 } 266 267 (void) VOP_BWRITE(bp); /* Ifile */ 268 lfs_vunref(vp); 269 270 return 0; 271 } 272 273 /* Allocate a new inode. */ 274 /* ARGSUSED */ 275 /* VOP_BWRITE 2i times */ 276 int 277 lfs_valloc(void *v) 278 { 279 struct vop_valloc_args /* { 280 struct vnode *a_pvp; 281 int a_mode; 282 struct ucred *a_cred; 283 struct vnode **a_vpp; 284 } */ *ap = v; 285 struct lfs *fs; 286 struct buf *bp, *cbp; 287 struct ifile *ifp; 288 ino_t new_ino; 289 int error; 290 int new_gen; 291 CLEANERINFO *cip; 292 293 fs = VTOI(ap->a_pvp)->i_lfs; 294 if (fs->lfs_ronly) 295 return EROFS; 296 *ap->a_vpp = NULL; 297 298 if (fs->lfs_version == 1) { 299 /* 300 * Use lfs_seglock here, instead of fs->lfs_freelock, to 301 * ensure that the free list is not changed in between 302 * the time that the ifile blocks are written to disk 303 * and the time that the superblock is written to disk. 304 */ 305 lfs_seglock(fs, SEGM_PROT); 306 } else { 307 lockmgr(&fs->lfs_freelock, LK_EXCLUSIVE, 0); 308 } 309 310 /* Get the head of the freelist. */ 311 LFS_GET_HEADFREE(fs, cip, cbp, &new_ino); 312 313 #ifdef DIAGNOSTIC 314 if(new_ino == LFS_UNUSED_INUM) { 315 #ifdef DEBUG 316 lfs_dump_super(fs); 317 #endif /* DEBUG */ 318 panic("inode 0 allocated [1]"); 319 } 320 #endif /* DIAGNOSTIC */ 321 #ifdef ALLOCPRINT 322 printf("lfs_valloc: allocate inode %d\n", new_ino); 323 #endif 324 325 /* 326 * Remove the inode from the free list and write the new start 327 * of the free list into the superblock. 328 */ 329 LFS_IENTRY(ifp, fs, new_ino, bp); 330 if (ifp->if_daddr != LFS_UNUSED_DADDR) 331 panic("lfs_valloc: inuse inode %d on the free list", new_ino); 332 LFS_PUT_HEADFREE(fs, cip, cbp, ifp->if_nextfree); 333 334 new_gen = ifp->if_version; /* version was updated by vfree */ 335 brelse(bp); 336 337 /* Extend IFILE so that the next lfs_valloc will succeed. */ 338 if (fs->lfs_free == LFS_UNUSED_INUM) { 339 if ((error = extend_ifile(fs, ap->a_cred)) != 0) { 340 LFS_PUT_HEADFREE(fs, cip, cbp, new_ino); 341 if (fs->lfs_version == 1) 342 lfs_segunlock(fs); 343 else 344 lockmgr(&fs->lfs_freelock, LK_RELEASE, 0); 345 return error; 346 } 347 } 348 #ifdef DIAGNOSTIC 349 if(fs->lfs_free == LFS_UNUSED_INUM) 350 panic("inode 0 allocated [3]"); 351 #endif /* DIAGNOSTIC */ 352 353 if (fs->lfs_version == 1) 354 lfs_segunlock(fs); 355 else 356 lockmgr(&fs->lfs_freelock, LK_RELEASE, 0); 357 358 return lfs_ialloc(fs, ap->a_pvp, new_ino, new_gen, ap->a_vpp); 359 } 360 361 static int 362 lfs_ialloc(struct lfs *fs, struct vnode *pvp, ino_t new_ino, int new_gen, 363 struct vnode **vpp) 364 { 365 struct inode *ip; 366 struct vnode *vp; 367 IFILE *ifp; 368 struct buf *bp, *cbp; 369 int error; 370 CLEANERINFO *cip; 371 372 error = getnewvnode(VT_LFS, pvp->v_mount, lfs_vnodeop_p, &vp); 373 /* printf("lfs_ialloc: ino %d vp %p error %d\n", new_ino, vp, error);*/ 374 if (error) 375 goto errout; 376 377 lockmgr(&ufs_hashlock, LK_EXCLUSIVE, 0); 378 /* Create an inode to associate with the vnode. */ 379 lfs_vcreate(pvp->v_mount, new_ino, vp); 380 381 ip = VTOI(vp); 382 LFS_SET_UINO(ip, IN_CHANGE | IN_MODIFIED); 383 /* Zero out the direct and indirect block addresses. */ 384 bzero(&ip->i_din, sizeof(ip->i_din)); 385 ip->i_din.ffs_din.di_inumber = new_ino; 386 387 /* Set a new generation number for this inode. */ 388 if (new_gen) 389 ip->i_ffs_gen = new_gen; 390 391 /* Insert into the inode hash table. */ 392 ufs_ihashins(ip); 393 lockmgr(&ufs_hashlock, LK_RELEASE, 0); 394 395 error = ufs_vinit(vp->v_mount, lfs_specop_p, lfs_fifoop_p, &vp); 396 ip = VTOI(vp); 397 if (error) { 398 vput(vp); 399 goto errout; 400 } 401 /* printf("lfs_ialloc[2]: ino %d vp %p\n", new_ino, vp);*/ 402 403 *vpp = vp; 404 #if 1 405 if(!(vp->v_flag & VDIROP)) { 406 (void)lfs_vref(vp); 407 ++lfs_dirvcount; 408 } 409 vp->v_flag |= VDIROP; 410 411 if(!(ip->i_flag & IN_ADIROP)) 412 ++fs->lfs_nadirop; 413 ip->i_flag |= IN_ADIROP; 414 #endif 415 VREF(ip->i_devvp); 416 /* Set superblock modified bit and increment file count. */ 417 fs->lfs_fmod = 1; 418 ++fs->lfs_nfiles; 419 return (0); 420 421 errout: 422 /* 423 * Put the new inum back on the free list. 424 */ 425 LFS_IENTRY(ifp, fs, new_ino, bp); 426 ifp->if_daddr = LFS_UNUSED_DADDR; 427 LFS_GET_HEADFREE(fs, cip, cbp, &(ifp->if_nextfree)); 428 LFS_PUT_HEADFREE(fs, cip, cbp, new_ino); 429 (void) VOP_BWRITE(bp); /* Ifile */ 430 431 *vpp = NULLVP; 432 return (error); 433 } 434 435 /* Create a new vnode/inode pair and initialize what fields we can. */ 436 void 437 lfs_vcreate(struct mount *mp, ino_t ino, struct vnode *vp) 438 { 439 struct inode *ip; 440 struct ufsmount *ump; 441 #ifdef QUOTA 442 int i; 443 #endif 444 445 /* Get a pointer to the private mount structure. */ 446 ump = VFSTOUFS(mp); 447 448 /* Initialize the inode. */ 449 ip = pool_get(&lfs_inode_pool, PR_WAITOK); 450 vp->v_data = ip; 451 ip->i_vnode = vp; 452 ip->i_devvp = ump->um_devvp; 453 ip->i_dev = ump->um_dev; 454 ip->i_number = ip->i_din.ffs_din.di_inumber = ino; 455 ip->i_lfs = ump->um_lfs; 456 #ifdef QUOTA 457 for (i = 0; i < MAXQUOTAS; i++) 458 ip->i_dquot[i] = NODQUOT; 459 #endif 460 ip->i_lockf = 0; 461 ip->i_diroff = 0; 462 ip->i_ffs_mode = 0; 463 ip->i_ffs_size = 0; 464 ip->i_ffs_blocks = 0; 465 ip->i_lfs_effnblks = 0; 466 ip->i_flag = 0; 467 /* Why was IN_MODIFIED ever set here? */ 468 /* LFS_SET_UINO(ip, IN_CHANGE | IN_MODIFIED); */ 469 } 470 471 /* Free an inode. */ 472 /* ARGUSED */ 473 /* VOP_BWRITE 2i times */ 474 int 475 lfs_vfree(void *v) 476 { 477 struct vop_vfree_args /* { 478 struct vnode *a_pvp; 479 ino_t a_ino; 480 int a_mode; 481 } */ *ap = v; 482 SEGUSE *sup; 483 CLEANERINFO *cip; 484 struct buf *cbp, *bp; 485 struct ifile *ifp; 486 struct inode *ip; 487 struct vnode *vp; 488 struct lfs *fs; 489 ufs_daddr_t old_iaddr; 490 ino_t ino, otail; 491 extern int lfs_dirvcount; 492 493 /* Get the inode number and file system. */ 494 vp = ap->a_pvp; 495 ip = VTOI(vp); 496 fs = ip->i_lfs; 497 ino = ip->i_number; 498 499 /* Drain of pending writes */ 500 if (fs->lfs_version > 1 && WRITEINPROG(vp)) 501 tsleep(vp, (PRIBIO+1), "lfs_vfree", 0); 502 503 if (fs->lfs_version == 1) 504 lfs_seglock(fs, SEGM_PROT); 505 else 506 lockmgr(&fs->lfs_freelock, LK_EXCLUSIVE, 0); 507 508 if(vp->v_flag & VDIROP) { 509 --lfs_dirvcount; 510 vp->v_flag &= ~VDIROP; 511 wakeup(&lfs_dirvcount); 512 lfs_vunref(vp); 513 } 514 if (ip->i_flag & IN_ADIROP) { 515 --fs->lfs_nadirop; 516 ip->i_flag &= ~IN_ADIROP; 517 } 518 519 LFS_CLR_UINO(ip, IN_ACCESSED|IN_CLEANING|IN_MODIFIED); 520 ip->i_flag &= ~IN_ALLMOD; 521 522 /* 523 * Set the ifile's inode entry to unused, increment its version number 524 * and link it onto the free chain. 525 */ 526 LFS_IENTRY(ifp, fs, ino, bp); 527 old_iaddr = ifp->if_daddr; 528 ifp->if_daddr = LFS_UNUSED_DADDR; 529 ++ifp->if_version; 530 if (fs->lfs_version == 1) { 531 LFS_GET_HEADFREE(fs, cip, cbp, &(ifp->if_nextfree)); 532 LFS_PUT_HEADFREE(fs, cip, cbp, ino); 533 (void) VOP_BWRITE(bp); /* Ifile */ 534 } else { 535 ifp->if_nextfree = LFS_UNUSED_INUM; 536 /* 537 * XXX Writing the freed node here means that it might not 538 * XXX make it into the free list in the event of a crash 539 * XXX (the ifile could be written before the rest of this 540 * XXX completes). 541 */ 542 (void) VOP_BWRITE(bp); /* Ifile */ 543 LFS_GET_TAILFREE(fs, cip, cbp, &otail); 544 LFS_IENTRY(ifp, fs, otail, bp); 545 ifp->if_nextfree = ino; 546 VOP_BWRITE(bp); 547 LFS_PUT_TAILFREE(fs, cip, cbp, ino); 548 } 549 #ifdef DIAGNOSTIC 550 if(ino == LFS_UNUSED_INUM) { 551 panic("inode 0 freed"); 552 } 553 #endif /* DIAGNOSTIC */ 554 if (old_iaddr != LFS_UNUSED_DADDR) { 555 LFS_SEGENTRY(sup, fs, dtosn(fs, old_iaddr), bp); 556 #ifdef DIAGNOSTIC 557 if (sup->su_nbytes < DINODE_SIZE) { 558 printf("lfs_vfree: negative byte count" 559 " (segment %d short by %d)\n", 560 dtosn(fs, old_iaddr), 561 (int)DINODE_SIZE - sup->su_nbytes); 562 panic("lfs_vfree: negative byte count"); 563 sup->su_nbytes = DINODE_SIZE; 564 } 565 #endif 566 sup->su_nbytes -= DINODE_SIZE; 567 (void) VOP_BWRITE(bp); /* Ifile */ 568 } 569 570 /* Set superblock modified bit and decrement file count. */ 571 fs->lfs_fmod = 1; 572 --fs->lfs_nfiles; 573 574 if (fs->lfs_version == 1) 575 lfs_segunlock(fs); 576 else 577 lockmgr(&fs->lfs_freelock, LK_RELEASE, 0); 578 return (0); 579 } 580