lfs_alloc.c revision 1.63
1 /* $NetBSD: lfs_alloc.c,v 1.63 2003/02/17 23:48:16 perseant Exp $ */ 2 3 /*- 4 * Copyright (c) 1999, 2000, 2001, 2002, 2003 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 (at) 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 1.63 2003/02/17 23:48:16 perseant 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/lock.h> 85 #include <sys/vnode.h> 86 #include <sys/syslog.h> 87 #include <sys/mount.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 extern struct simplelock lfs_subsys_lock; 102 extern int lfs_subsys_pages; 103 104 static int extend_ifile(struct lfs *, struct ucred *); 105 static int lfs_ialloc(struct lfs *, struct vnode *, ino_t, int, struct vnode **); 106 107 /* 108 * Allocate a particular inode with a particular version number, freeing 109 * any previous versions of this inode that may have gone before. 110 * Used by the roll-forward code. 111 * 112 * XXX this function does not have appropriate locking to be used on a live fs; 113 * XXX but something similar could probably be used for an "undelete" call. 114 * 115 * Called with the Ifile inode locked. 116 */ 117 int 118 lfs_rf_valloc(struct lfs *fs, ino_t ino, int version, struct proc *p, 119 struct vnode **vpp) 120 { 121 IFILE *ifp; 122 struct buf *bp, *cbp; 123 struct vnode *vp; 124 struct inode *ip; 125 ino_t tino, oldnext; 126 int error; 127 CLEANERINFO *cip; 128 129 /* 130 * First, just try a vget. If the version number is the one we want, 131 * we don't have to do anything else. If the version number is wrong, 132 * take appropriate action. 133 */ 134 error = VFS_VGET(fs->lfs_ivnode->v_mount, ino, &vp); 135 if (error == 0) { 136 /* printf("lfs_rf_valloc[1]: ino %d vp %p\n", ino, vp); */ 137 138 *vpp = vp; 139 ip = VTOI(vp); 140 if (ip->i_ffs_gen == version) 141 return 0; 142 else if (ip->i_ffs_gen < version) { 143 VOP_TRUNCATE(vp, (off_t)0, 0, NOCRED, p); 144 ip->i_ffs_gen = version; 145 LFS_SET_UINO(ip, IN_CHANGE | IN_MODIFIED | IN_UPDATE); 146 return 0; 147 } else { 148 /* printf("ino %d: asked for version %d but got %d\n", 149 ino, version, ip->i_ffs_gen); */ 150 vput(vp); 151 *vpp = NULLVP; 152 return EEXIST; 153 } 154 } 155 156 /* 157 * The inode is not in use. Find it on the free list. 158 */ 159 /* If the Ifile is too short to contain this inum, extend it */ 160 while (VTOI(fs->lfs_ivnode)->i_ffs_size <= (ino / 161 fs->lfs_ifpb + fs->lfs_cleansz + fs->lfs_segtabsz) 162 << fs->lfs_bshift) { 163 extend_ifile(fs, NOCRED); 164 } 165 166 LFS_IENTRY(ifp, fs, ino, bp); 167 oldnext = ifp->if_nextfree; 168 ifp->if_version = version; 169 brelse(bp); 170 171 LFS_GET_HEADFREE(fs, cip, cbp, &ino); 172 if (ino) { 173 LFS_PUT_HEADFREE(fs, cip, cbp, oldnext); 174 } else { 175 tino = ino; 176 while (1) { 177 LFS_IENTRY(ifp, fs, tino, bp); 178 if (ifp->if_nextfree == ino || 179 ifp->if_nextfree == LFS_UNUSED_INUM) 180 break; 181 tino = ifp->if_nextfree; 182 brelse(bp); 183 } 184 if (ifp->if_nextfree == LFS_UNUSED_INUM) { 185 brelse(bp); 186 return ENOENT; 187 } 188 ifp->if_nextfree = oldnext; 189 LFS_BWRITE_LOG(bp); 190 } 191 192 error = lfs_ialloc(fs, fs->lfs_ivnode, ino, version, &vp); 193 if (error == 0) { 194 /* 195 * Make it VREG so we can put blocks on it. We will change 196 * this later if it turns out to be some other kind of file. 197 */ 198 ip = VTOI(vp); 199 ip->i_ffs_mode = IFREG; 200 ip->i_ffs_nlink = 1; 201 ip->i_ffs_effnlink = 1; 202 ufs_vinit(vp->v_mount, lfs_specop_p, lfs_fifoop_p, &vp); 203 ip = VTOI(vp); 204 205 /* printf("lfs_rf_valloc: ino %d vp %p\n", ino, vp); */ 206 207 /* The dirop-nature of this vnode is past */ 208 (void)lfs_vunref(vp); 209 --lfs_dirvcount; 210 vp->v_flag &= ~VDIROP; 211 TAILQ_REMOVE(&fs->lfs_dchainhd, ip, i_lfs_dchain); 212 --fs->lfs_nadirop; 213 ip->i_flag &= ~IN_ADIROP; 214 } 215 *vpp = vp; 216 return error; 217 } 218 219 /* 220 * Called with the Ifile inode locked. 221 */ 222 static int 223 extend_ifile(struct lfs *fs, struct ucred *cred) 224 { 225 struct vnode *vp; 226 struct inode *ip; 227 IFILE *ifp; 228 IFILE_V1 *ifp_v1; 229 struct buf *bp, *cbp; 230 int error; 231 daddr_t i, blkno, max; 232 ino_t oldlast; 233 CLEANERINFO *cip; 234 235 vp = fs->lfs_ivnode; 236 ip = VTOI(vp); 237 blkno = lblkno(fs, ip->i_ffs_size); 238 if ((error = VOP_BALLOC(vp, ip->i_ffs_size, fs->lfs_bsize, cred, 0, 239 &bp)) != 0) { 240 return (error); 241 } 242 ip->i_ffs_size += fs->lfs_bsize; 243 uvm_vnp_setsize(vp, ip->i_ffs_size); 244 245 i = (blkno - fs->lfs_segtabsz - fs->lfs_cleansz) * 246 fs->lfs_ifpb; 247 LFS_GET_HEADFREE(fs, cip, cbp, &oldlast); 248 LFS_PUT_HEADFREE(fs, cip, cbp, i); 249 #ifdef DIAGNOSTIC 250 if (fs->lfs_freehd == LFS_UNUSED_INUM) 251 panic("inode 0 allocated [2]"); 252 #endif /* DIAGNOSTIC */ 253 max = i + fs->lfs_ifpb; 254 /* printf("extend_ifile: new block ino %d--%d\n", i, max - 1); */ 255 256 if (fs->lfs_version == 1) { 257 for (ifp_v1 = (IFILE_V1 *)bp->b_data; i < max; ++ifp_v1) { 258 ifp_v1->if_version = 1; 259 ifp_v1->if_daddr = LFS_UNUSED_DADDR; 260 ifp_v1->if_nextfree = ++i; 261 } 262 ifp_v1--; 263 ifp_v1->if_nextfree = oldlast; 264 } else { 265 for (ifp = (IFILE *)bp->b_data; i < max; ++ifp) { 266 ifp->if_version = 1; 267 ifp->if_daddr = LFS_UNUSED_DADDR; 268 ifp->if_nextfree = ++i; 269 } 270 ifp--; 271 ifp->if_nextfree = oldlast; 272 } 273 LFS_PUT_TAILFREE(fs, cip, cbp, max - 1); 274 275 (void) LFS_BWRITE_LOG(bp); /* Ifile */ 276 277 return 0; 278 } 279 280 /* Allocate a new inode. */ 281 /* ARGSUSED */ 282 /* VOP_BWRITE 2i times */ 283 int 284 lfs_valloc(void *v) 285 { 286 struct vop_valloc_args /* { 287 struct vnode *a_pvp; 288 int a_mode; 289 struct ucred *a_cred; 290 struct vnode **a_vpp; 291 } */ *ap = v; 292 struct lfs *fs; 293 struct buf *bp, *cbp; 294 struct ifile *ifp; 295 ino_t new_ino; 296 int error; 297 int new_gen; 298 CLEANERINFO *cip; 299 300 fs = VTOI(ap->a_pvp)->i_lfs; 301 if (fs->lfs_ronly) 302 return EROFS; 303 *ap->a_vpp = NULL; 304 305 lfs_seglock(fs, SEGM_PROT); 306 307 /* Get the head of the freelist. */ 308 LFS_GET_HEADFREE(fs, cip, cbp, &new_ino); 309 310 #ifdef DIAGNOSTIC 311 if (new_ino == LFS_UNUSED_INUM) { 312 #ifdef DEBUG 313 lfs_dump_super(fs); 314 #endif /* DEBUG */ 315 panic("inode 0 allocated [1]"); 316 } 317 #endif /* DIAGNOSTIC */ 318 #ifdef ALLOCPRINT 319 printf("lfs_valloc: allocate inode %d\n", new_ino); 320 #endif 321 322 /* 323 * Remove the inode from the free list and write the new start 324 * of the free list into the superblock. 325 */ 326 LFS_IENTRY(ifp, fs, new_ino, bp); 327 if (ifp->if_daddr != LFS_UNUSED_DADDR) 328 panic("lfs_valloc: inuse inode %d on the free list", new_ino); 329 LFS_PUT_HEADFREE(fs, cip, cbp, ifp->if_nextfree); 330 /* printf("lfs_valloc: headfree %d -> %d\n", new_ino, ifp->if_nextfree); */ 331 332 new_gen = ifp->if_version; /* version was updated by vfree */ 333 brelse(bp); 334 335 /* Extend IFILE so that the next lfs_valloc will succeed. */ 336 if (fs->lfs_freehd == LFS_UNUSED_INUM) { 337 if ((error = extend_ifile(fs, ap->a_cred)) != 0) { 338 LFS_PUT_HEADFREE(fs, cip, cbp, new_ino); 339 lfs_segunlock(fs); 340 return error; 341 } 342 } 343 #ifdef DIAGNOSTIC 344 if (fs->lfs_freehd == LFS_UNUSED_INUM) 345 panic("inode 0 allocated [3]"); 346 #endif /* DIAGNOSTIC */ 347 348 lfs_segunlock(fs); 349 350 return lfs_ialloc(fs, ap->a_pvp, new_ino, new_gen, ap->a_vpp); 351 } 352 353 static int 354 lfs_ialloc(struct lfs *fs, struct vnode *pvp, ino_t new_ino, int new_gen, 355 struct vnode **vpp) 356 { 357 struct inode *ip; 358 struct vnode *vp; 359 IFILE *ifp; 360 struct buf *bp, *cbp; 361 int error; 362 CLEANERINFO *cip; 363 364 error = getnewvnode(VT_LFS, pvp->v_mount, lfs_vnodeop_p, &vp); 365 /* printf("lfs_ialloc: ino %d vp %p error %d\n", new_ino, vp, error);*/ 366 if (error) 367 goto errout; 368 369 lockmgr(&ufs_hashlock, LK_EXCLUSIVE, 0); 370 /* Create an inode to associate with the vnode. */ 371 lfs_vcreate(pvp->v_mount, new_ino, vp); 372 373 ip = VTOI(vp); 374 LFS_SET_UINO(ip, IN_CHANGE | IN_MODIFIED); 375 /* Zero out the direct and indirect block addresses. */ 376 bzero(&ip->i_din, sizeof(ip->i_din)); 377 ip->i_din.ffs_din.di_inumber = new_ino; 378 379 /* Set a new generation number for this inode. */ 380 if (new_gen) 381 ip->i_ffs_gen = new_gen; 382 383 /* Insert into the inode hash table. */ 384 ufs_ihashins(ip); 385 lockmgr(&ufs_hashlock, LK_RELEASE, 0); 386 387 ufs_vinit(vp->v_mount, lfs_specop_p, lfs_fifoop_p, &vp); 388 ip = VTOI(vp); 389 /* printf("lfs_ialloc[2]: ino %d vp %p\n", new_ino, vp);*/ 390 391 memset(ip->i_lfs_fragsize, 0, NDADDR * sizeof(*ip->i_lfs_fragsize)); 392 393 uvm_vnp_setsize(vp, 0); 394 *vpp = vp; 395 if (!(vp->v_flag & VDIROP)) { 396 (void)lfs_vref(vp); 397 ++lfs_dirvcount; 398 TAILQ_INSERT_TAIL(&fs->lfs_dchainhd, ip, i_lfs_dchain); 399 } 400 vp->v_flag |= VDIROP; 401 402 if (!(ip->i_flag & IN_ADIROP)) 403 ++fs->lfs_nadirop; 404 ip->i_flag |= IN_ADIROP; 405 genfs_node_init(vp, &lfs_genfsops); 406 VREF(ip->i_devvp); 407 /* Set superblock modified bit and increment file count. */ 408 fs->lfs_fmod = 1; 409 ++fs->lfs_nfiles; 410 return (0); 411 412 errout: 413 /* 414 * Put the new inum back on the free list. 415 */ 416 lfs_seglock(fs, SEGM_PROT); 417 LFS_IENTRY(ifp, fs, new_ino, bp); 418 ifp->if_daddr = LFS_UNUSED_DADDR; 419 LFS_GET_HEADFREE(fs, cip, cbp, &(ifp->if_nextfree)); 420 LFS_PUT_HEADFREE(fs, cip, cbp, new_ino); 421 (void) LFS_BWRITE_LOG(bp); /* Ifile */ 422 lfs_segunlock(fs); 423 424 *vpp = NULLVP; 425 return (error); 426 } 427 428 /* Create a new vnode/inode pair and initialize what fields we can. */ 429 void 430 lfs_vcreate(struct mount *mp, ino_t ino, struct vnode *vp) 431 { 432 struct inode *ip; 433 struct ufsmount *ump; 434 #ifdef QUOTA 435 int i; 436 #endif 437 438 /* Get a pointer to the private mount structure. */ 439 ump = VFSTOUFS(mp); 440 441 /* Initialize the inode. */ 442 ip = pool_get(&lfs_inode_pool, PR_WAITOK); 443 ip->inode_ext.lfs = pool_get(&lfs_inoext_pool, PR_WAITOK); 444 vp->v_data = ip; 445 ip->i_vnode = vp; 446 ip->i_devvp = ump->um_devvp; 447 ip->i_dev = ump->um_dev; 448 ip->i_number = ip->i_din.ffs_din.di_inumber = ino; 449 ip->i_lfs = ump->um_lfs; 450 #ifdef QUOTA 451 for (i = 0; i < MAXQUOTAS; i++) 452 ip->i_dquot[i] = NODQUOT; 453 #endif 454 ip->i_lockf = 0; 455 ip->i_diroff = 0; 456 ip->i_ffs_mode = 0; 457 ip->i_ffs_size = 0; 458 ip->i_ffs_blocks = 0; 459 ip->i_lfs_effnblks = 0; 460 ip->i_flag = 0; 461 462 #ifdef DEBUG_LFS_VNLOCK 463 if (ino == LFS_IFILE_INUM) 464 vp->v_vnlock->lk_wmesg = "inlock"; 465 #endif 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 daddr_t old_iaddr; 487 ino_t ino, otail; 488 extern int lfs_dirvcount; 489 int s; 490 491 /* Get the inode number and file system. */ 492 vp = ap->a_pvp; 493 ip = VTOI(vp); 494 fs = ip->i_lfs; 495 ino = ip->i_number; 496 497 /* Drain of pending writes */ 498 s = splbio(); 499 if (fs->lfs_version > 1 && WRITEINPROG(vp)) 500 tsleep(vp, (PRIBIO+1), "lfs_vfree", 0); 501 splx(s); 502 503 lfs_seglock(fs, SEGM_PROT); 504 505 if (vp->v_flag & VDIROP) { 506 --lfs_dirvcount; 507 vp->v_flag &= ~VDIROP; 508 TAILQ_REMOVE(&fs->lfs_dchainhd, ip, i_lfs_dchain); 509 wakeup(&lfs_dirvcount); 510 lfs_vunref(vp); 511 } 512 lfs_unmark_vnode(vp); 513 514 LFS_CLR_UINO(ip, IN_ACCESSED|IN_CLEANING|IN_MODIFIED); 515 ip->i_flag &= ~IN_ALLMOD; 516 517 /* 518 * Set the ifile's inode entry to unused, increment its version number 519 * and link it onto the free chain. 520 */ 521 LFS_IENTRY(ifp, fs, ino, bp); 522 old_iaddr = ifp->if_daddr; 523 ifp->if_daddr = LFS_UNUSED_DADDR; 524 ++ifp->if_version; 525 if (fs->lfs_version == 1) { 526 LFS_GET_HEADFREE(fs, cip, cbp, &(ifp->if_nextfree)); 527 LFS_PUT_HEADFREE(fs, cip, cbp, ino); 528 (void) LFS_BWRITE_LOG(bp); /* Ifile */ 529 } else { 530 ifp->if_nextfree = LFS_UNUSED_INUM; 531 /* 532 * XXX Writing the freed node here means that it might not 533 * XXX make it into the free list in the event of a crash 534 * XXX (the ifile could be written before the rest of this 535 * XXX completes). 536 */ 537 (void) LFS_BWRITE_LOG(bp); /* Ifile */ 538 LFS_GET_TAILFREE(fs, cip, cbp, &otail); 539 LFS_IENTRY(ifp, fs, otail, bp); 540 ifp->if_nextfree = ino; 541 LFS_BWRITE_LOG(bp); 542 LFS_PUT_TAILFREE(fs, cip, cbp, ino); 543 /* printf("lfs_vfree: tailfree %d -> %d\n", otail, ino); */ 544 } 545 #ifdef DIAGNOSTIC 546 if (ino == LFS_UNUSED_INUM) { 547 panic("inode 0 freed"); 548 } 549 #endif /* DIAGNOSTIC */ 550 if (old_iaddr != LFS_UNUSED_DADDR) { 551 LFS_SEGENTRY(sup, fs, dtosn(fs, old_iaddr), bp); 552 #ifdef DIAGNOSTIC 553 if (sup->su_nbytes < DINODE_SIZE) { 554 printf("lfs_vfree: negative byte count" 555 " (segment %" PRIu32 " short by %d)\n", 556 dtosn(fs, old_iaddr), 557 (int)DINODE_SIZE - sup->su_nbytes); 558 panic("lfs_vfree: negative byte count"); 559 sup->su_nbytes = DINODE_SIZE; 560 } 561 #endif 562 sup->su_nbytes -= DINODE_SIZE; 563 LFS_WRITESEGENTRY(sup, fs, dtosn(fs, old_iaddr), bp); /* Ifile */ 564 } 565 566 /* Set superblock modified bit and decrement file count. */ 567 fs->lfs_fmod = 1; 568 --fs->lfs_nfiles; 569 570 lfs_segunlock(fs); 571 572 return (0); 573 } 574
Indexes created Fri Sep 26 20:09:58 GMT 2025