lfs_inode.c revision 1.22.2.1.2.3
1 /* $NetBSD: lfs_inode.c,v 1.22.2.1.2.3 1999/08/31 21:03:46 perseant Exp $ */ 2 3 /*- 4 * Copyright (c) 1999 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) 1986, 1989, 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_inode.c 8.9 (Berkeley) 5/8/95 71 */ 72 73 #if defined(_KERNEL) && !defined(_LKM) 74 #include "opt_quota.h" 75 #endif 76 77 #include <sys/param.h> 78 #include <sys/systm.h> 79 #include <sys/mount.h> 80 #include <sys/proc.h> 81 #include <sys/file.h> 82 #include <sys/buf.h> 83 #include <sys/vnode.h> 84 #include <sys/kernel.h> 85 #include <sys/malloc.h> 86 87 #include <vm/vm.h> 88 89 #include <ufs/ufs/quota.h> 90 #include <ufs/ufs/inode.h> 91 #include <ufs/ufs/ufsmount.h> 92 #include <ufs/ufs/ufs_extern.h> 93 94 #include <ufs/lfs/lfs.h> 95 #include <ufs/lfs/lfs_extern.h> 96 97 static int lfs_vinvalbuf __P((struct vnode *, struct ucred *, struct proc *, ufs_daddr_t)); 98 99 /* Search a block for a specific dinode. */ 100 struct dinode * 101 lfs_ifind(fs, ino, dip) 102 struct lfs *fs; 103 ino_t ino; 104 register struct dinode *dip; 105 { 106 register int cnt; 107 register struct dinode *ldip; 108 109 for (cnt = INOPB(fs), ldip = dip + (cnt - 1); cnt--; --ldip) 110 if (ldip->di_inumber == ino) 111 return (ldip); 112 113 panic("lfs_ifind: dinode %u not found", ino); 114 /* NOTREACHED */ 115 } 116 117 int 118 lfs_update(v) 119 void *v; 120 { 121 struct vop_update_args /* { 122 struct vnode *a_vp; 123 struct timespec *a_access; 124 struct timespec *a_modify; 125 int a_waitfor; 126 } */ *ap = v; 127 struct inode *ip; 128 struct vnode *vp = ap->a_vp; 129 int mod, oflag; 130 struct timespec ts; 131 struct lfs *fs = VFSTOUFS(vp->v_mount)->um_lfs; 132 133 if (vp->v_mount->mnt_flag & MNT_RDONLY) 134 return (0); 135 ip = VTOI(vp); 136 137 /* 138 * If we are called from vinvalbuf, and the file's blocks have 139 * already been scheduled for writing, but the writes have not 140 * yet completed, lfs_vflush will not be called, and vinvalbuf 141 * will cause a panic. So, we must wait until any pending write 142 * for our inode completes, if we are called with LFS_SYNC set. 143 */ 144 while((ap->a_waitfor & LFS_SYNC) && WRITEINPROG(vp)) { 145 #ifdef DEBUG_LFS 146 printf("lfs_update: sleeping on inode %d (in-progress)\n",ip->i_number); 147 #endif 148 tsleep(vp, (PRIBIO+1), "lfs_update", 0); 149 } 150 mod = ip->i_flag & IN_MODIFIED; 151 oflag = ip->i_flag; 152 TIMEVAL_TO_TIMESPEC(&time, &ts); 153 LFS_ITIMES(ip, 154 ap->a_access ? ap->a_access : &ts, 155 ap->a_modify ? ap->a_modify : &ts, &ts); 156 if (!mod && (ip->i_flag & IN_MODIFIED)) 157 ip->i_lfs->lfs_uinodes++; 158 if ((ip->i_flag & (IN_MODIFIED|IN_CLEANING)) == 0) { 159 return (0); 160 } 161 162 /* If sync, push back the vnode and any dirty blocks it may have. */ 163 if(ap->a_waitfor & LFS_SYNC) { 164 /* Avoid flushing VDIROP. */ 165 ++fs->lfs_diropwait; 166 while(vp->v_flag & VDIROP) { 167 #ifdef DEBUG_LFS 168 printf("lfs_update: sleeping on inode %d (dirops)\n",ip->i_number); 169 #endif 170 if(fs->lfs_dirops==0) 171 lfs_flush_fs(vp->v_mount,0); 172 else 173 tsleep(&fs->lfs_writer, PRIBIO+1, "lfs_fsync", 0); 174 /* XXX KS - by falling out here, are we writing the vn 175 twice? */ 176 } 177 --fs->lfs_diropwait; 178 return lfs_vflush(vp); 179 } 180 return 0; 181 } 182 183 /* Update segment usage information when removing a block. */ 184 #define UPDATE_SEGUSE do {\ 185 if (lastseg != -1) { \ 186 LFS_SEGENTRY(sup, fs, lastseg, sup_bp); \ 187 if (num > sup->su_nbytes) { \ 188 printf("lfs_truncate: negative bytes: segment %d short by %d\n", \ 189 lastseg, num - sup->su_nbytes); \ 190 panic("lfs_truncate: negative bytes"); \ 191 sup->su_nbytes = 0; \ 192 } else \ 193 sup->su_nbytes -= num; \ 194 e1 = VOP_BWRITE(sup_bp); \ 195 fragsreleased += numfrags(fs, num); \ 196 } \ 197 fragsreleased += numfrags(fs, unum); \ 198 } while(0) 199 200 #define SEGDEC(S) do { \ 201 if (daddr > 0) { \ 202 if (lastseg != (seg = datosn(fs, daddr))) { \ 203 UPDATE_SEGUSE; \ 204 num = (S); \ 205 lastseg = seg; \ 206 } else \ 207 num += (S); \ 208 } else if (daddr == UNWRITTEN) { \ 209 unum += (S); \ 210 } \ 211 } while(0) 212 213 /* 214 * Truncate the inode ip to at most length size. Update segment usage 215 * table information. 216 */ 217 /* ARGSUSED */ 218 int 219 lfs_truncate(v) 220 void *v; 221 { 222 struct vop_truncate_args /* { 223 struct vnode *a_vp; 224 off_t a_length; 225 int a_flags; 226 struct ucred *a_cred; 227 struct proc *a_p; 228 } */ *ap = v; 229 register struct indir *inp; 230 register int i; 231 register ufs_daddr_t *daddrp; 232 register struct vnode *vp = ap->a_vp; 233 off_t length = ap->a_length; 234 struct buf *bp, *sup_bp; 235 struct ifile *ifp; 236 struct inode *ip; 237 struct lfs *fs; 238 struct indir a[NIADDR + 2], a_end[NIADDR + 2]; 239 SEGUSE *sup; 240 ufs_daddr_t daddr, lastblock, lbn, olastblock; 241 ufs_daddr_t oldsize_lastblock, oldsize_newlast, newsize; 242 long off, a_released, fragsreleased, i_released; 243 int e1, e2, depth, lastseg, num, unum, offset, seg, freesize; 244 245 ip = VTOI(vp); 246 247 if (vp->v_type == VLNK && vp->v_mount->mnt_maxsymlinklen > 0) { 248 #ifdef DIAGNOSTIC 249 if (length != 0) 250 panic("lfs_truncate: partial truncate of symlink"); 251 #endif 252 bzero((char *)&ip->i_ffs_shortlink, (u_int)ip->i_ffs_size); 253 ip->i_ffs_size = 0; 254 ip->i_flag |= IN_CHANGE | IN_UPDATE; 255 return (VOP_UPDATE(vp, NULL, NULL, 0)); 256 } 257 uvm_vnp_setsize(vp, length); 258 259 fs = ip->i_lfs; 260 lfs_imtime(fs); 261 262 /* If length is larger than the file, just update the times. */ 263 if (ip->i_ffs_size <= length) { 264 ip->i_flag |= IN_CHANGE | IN_UPDATE; 265 return (VOP_UPDATE(vp, NULL, NULL, 0)); 266 } 267 268 /* 269 * Make sure no writes happen while we're truncating. 270 * Otherwise, blocks which are accounted for on the inode 271 * *and* which have been created for cleaning can coexist, 272 * and cause us to overcount, and panic below. 273 * 274 * XXX KS - too restrictive? Maybe only when cleaning? 275 */ 276 while(fs->lfs_seglock) { 277 tsleep(&fs->lfs_seglock, (PRIBIO+1), "lfs_truncate", 0); 278 } 279 280 /* 281 * Calculate index into inode's block list of last direct and indirect 282 * blocks (if any) which we want to keep. Lastblock is 0 when the 283 * file is truncated to 0. 284 */ 285 lastblock = lblkno(fs, length + fs->lfs_bsize - 1); 286 olastblock = lblkno(fs, ip->i_ffs_size + fs->lfs_bsize - 1) - 1; 287 288 /* 289 * Update the size of the file. If the file is not being truncated to 290 * a block boundry, the contents of the partial block following the end 291 * must be zero'd in case it ever becomes accessible again 292 * because of subsequent file growth. For this part of the code, 293 * oldsize_newlast refers to the old size of the new last block in the 294 * file. 295 */ 296 offset = blkoff(fs, length); 297 lbn = lblkno(fs, length); 298 oldsize_newlast = blksize(fs, ip, lbn); 299 300 /* Now set oldsize to the current size of the current last block */ 301 oldsize_lastblock = blksize(fs, ip, olastblock); 302 if (offset == 0) 303 ip->i_ffs_size = length; 304 else { 305 #ifdef QUOTA 306 if ((e1 = getinoquota(ip)) != 0) 307 return (e1); 308 #endif 309 if ((e1 = bread(vp, lbn, oldsize_newlast, NOCRED, &bp)) != 0) { 310 printf("lfs_truncate: bread: %d\n",e1); 311 brelse(bp); 312 return (e1); 313 } 314 ip->i_ffs_size = length; 315 newsize = blksize(fs, ip, lbn); 316 bzero((char *)bp->b_data + offset, (u_int)(newsize - offset)); 317 #ifdef DEBUG 318 if(lfs_iscleaner_bp(bp)) 319 panic("Can't allocbuf malloced buffer!"); 320 else 321 #endif 322 allocbuf(bp, newsize); 323 if(oldsize_newlast > newsize) 324 ip->i_ffs_blocks -= btodb(oldsize_newlast - newsize); 325 if ((e1 = VOP_BWRITE(bp)) != 0) { 326 printf("lfs_truncate: bwrite: %d\n",e1); 327 return (e1); 328 } 329 } 330 /* 331 * Modify sup->su_nbyte counters for each deleted block; keep track 332 * of number of blocks removed for ip->i_ffs_blocks. 333 */ 334 fragsreleased = 0; 335 num = 0; 336 unum = 0; 337 lastseg = -1; 338 339 for (lbn = olastblock; lbn >= lastblock;) { 340 /* XXX use run length from bmap array to make this faster */ 341 ufs_bmaparray(vp, lbn, &daddr, a, &depth, NULL); 342 if (lbn == olastblock) { 343 for (i = NIADDR + 2; i--;) 344 a_end[i] = a[i]; 345 freesize = oldsize_lastblock; 346 } else 347 freesize = fs->lfs_bsize; 348 switch (depth) { 349 case 0: /* Direct block. */ 350 daddr = ip->i_ffs_db[lbn]; 351 SEGDEC(freesize); 352 ip->i_ffs_db[lbn] = 0; 353 --lbn; 354 break; 355 #ifdef DIAGNOSTIC 356 case 1: /* An indirect block. */ 357 panic("lfs_truncate: ufs_bmaparray returned depth 1"); 358 /* NOTREACHED */ 359 #endif 360 default: /* Chain of indirect blocks. */ 361 inp = a + --depth; 362 if (inp->in_off > 0 && lbn != lastblock) { 363 lbn -= inp->in_off < lbn - lastblock ? 364 inp->in_off : lbn - lastblock; 365 break; 366 } 367 for (; depth && (inp->in_off == 0 || lbn == lastblock); 368 --inp, --depth) { 369 if (bread(vp, 370 inp->in_lbn, fs->lfs_bsize, NOCRED, &bp)) 371 panic("lfs_truncate: bread bno %d", 372 inp->in_lbn); 373 daddrp = (ufs_daddr_t *)bp->b_data + inp->in_off; 374 for (i = inp->in_off; 375 i++ <= a_end[depth].in_off;) { 376 daddr = *daddrp++; 377 SEGDEC(freesize); 378 } 379 a_end[depth].in_off = NINDIR(fs) - 1; 380 if (inp->in_off == 0) 381 brelse (bp); 382 else { 383 bzero((ufs_daddr_t *)bp->b_data + 384 inp->in_off, fs->lfs_bsize - 385 inp->in_off * sizeof(ufs_daddr_t)); 386 if ((e1 = VOP_BWRITE(bp)) != 0) { 387 printf("lfs_truncate: indir bwrite: %d\n",e1); 388 return (e1); 389 } 390 } 391 } 392 if (depth == 0 && a[1].in_off == 0) { 393 off = a[0].in_off; 394 daddr = ip->i_ffs_ib[off]; 395 SEGDEC(freesize); 396 ip->i_ffs_ib[off] = 0; 397 } 398 if (lbn == lastblock || lbn <= NDADDR) 399 --lbn; 400 else { 401 lbn -= NINDIR(fs); 402 if (lbn < lastblock) 403 lbn = lastblock; 404 } 405 } 406 } 407 UPDATE_SEGUSE; 408 409 /* If truncating the file to 0, update the version number. */ 410 if (length == 0) { 411 LFS_IENTRY(ifp, fs, ip->i_number, bp); 412 ++ifp->if_version; 413 (void) VOP_BWRITE(bp); 414 } 415 #ifdef DIAGNOSTIC 416 if (ip->i_ffs_blocks < fragstodb(fs, fragsreleased)) { 417 panic("lfs_truncate: frag count < 0 (%d<%ld), ino %d\n", 418 ip->i_ffs_blocks, fragstodb(fs, fragsreleased), 419 ip->i_number); 420 fragsreleased = dbtofrags(fs, ip->i_ffs_blocks); 421 } 422 #endif 423 ip->i_ffs_blocks -= fragstodb(fs, fragsreleased); 424 fs->lfs_bfree += fragstodb(fs, fragsreleased); 425 ip->i_flag |= IN_CHANGE | IN_UPDATE; 426 /* 427 * Traverse dirty block list counting number of dirty buffers 428 * that are being deleted out of the cache, so that the lfs_avail 429 * field can be updated. 430 */ 431 a_released = 0; 432 i_released = 0; 433 for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = bp->b_vnbufs.le_next) { 434 435 /* XXX KS - Don't miscount if we're not truncating to zero. */ 436 if(length>0 && !(bp->b_lblkno >= 0 && bp->b_lblkno > lastblock) 437 && !(bp->b_lblkno < 0 && bp->b_lblkno < -lastblock-NIADDR)) 438 continue; 439 440 if (bp->b_flags & B_LOCKED) { 441 a_released += numfrags(fs, bp->b_bcount); 442 /* 443 * XXX 444 * When buffers are created in the cache, their block 445 * number is set equal to their logical block number. 446 * If that is still true, we are assuming that the 447 * blocks are new (not yet on disk) and weren't 448 * counted above. However, there is a slight chance 449 * that a block's disk address is equal to its logical 450 * block number in which case, we'll get an overcounting 451 * here. 452 */ 453 if (bp->b_blkno == bp->b_lblkno) { 454 i_released += numfrags(fs, bp->b_bcount); 455 } 456 } 457 } 458 fragsreleased = i_released; 459 #ifdef DIAGNOSTIC 460 if (fragsreleased > dbtofrags(fs, ip->i_ffs_blocks)) { 461 printf("lfs_inode: %ld frags released > %d in inode %d\n", 462 fragsreleased, dbtofrags(fs, ip->i_ffs_blocks), 463 ip->i_number); 464 fragsreleased = dbtofrags(fs, ip->i_ffs_blocks); 465 } 466 #endif 467 fs->lfs_bfree += fragstodb(fs, fragsreleased); 468 ip->i_ffs_blocks -= fragstodb(fs, fragsreleased); 469 #ifdef DIAGNOSTIC 470 if (length == 0 && ip->i_ffs_blocks != 0) { 471 printf("lfs_inode: trunc to zero, but %d blocks left on inode %d\n", 472 ip->i_ffs_blocks, ip->i_number); 473 panic("lfs_inode\n"); 474 } 475 #endif 476 fs->lfs_avail += fragstodb(fs, a_released); 477 if(length>0) 478 e1 = lfs_vinvalbuf(vp, ap->a_cred, ap->a_p, lastblock-1); 479 else 480 e1 = vinvalbuf(vp, 0, ap->a_cred, ap->a_p, 0, 0); 481 e2 = VOP_UPDATE(vp, NULL, NULL, 0); 482 if(e1) 483 printf("lfs_truncate: vinvalbuf: %d\n",e1); 484 if(e2) 485 printf("lfs_truncate: update: %d\n",e2); 486 487 return (e1 ? e1 : e2 ? e2 : 0); 488 } 489 490 /* 491 * Get rid of blocks a la vinvalbuf; but only blocks that are of a higher 492 * lblkno than the file size allows. 493 */ 494 int 495 lfs_vinvalbuf(vp, cred, p, maxblk) 496 register struct vnode *vp; 497 struct ucred *cred; 498 struct proc *p; 499 ufs_daddr_t maxblk; 500 { 501 register struct buf *bp; 502 struct buf *nbp, *blist; 503 int i, s, error, dirty; 504 505 top: 506 dirty=0; 507 for (i=0;i<2;i++) { 508 if(i==0) 509 blist = vp->v_cleanblkhd.lh_first; 510 else /* i == 1 */ 511 blist = vp->v_dirtyblkhd.lh_first; 512 513 for (bp = blist; bp; bp = nbp) { 514 nbp = bp->b_vnbufs.le_next; 515 516 s = splbio(); 517 if (bp->b_flags & B_GATHERED) { 518 error = tsleep(vp, PRIBIO+1, "lfs_vin2", 0); 519 splx(s); 520 if(error) 521 return error; 522 goto top; 523 } 524 if (bp->b_flags & B_BUSY) { 525 bp->b_flags |= B_WANTED; 526 error = tsleep((caddr_t)bp, 527 (PRIBIO + 1), "lfs_vinval", 0); 528 splx(s); 529 if (error) 530 return (error); 531 goto top; 532 } 533 534 bp->b_flags |= B_BUSY; 535 splx(s); 536 if((bp->b_lblkno >= 0 && bp->b_lblkno > maxblk) 537 || (bp->b_lblkno < 0 && bp->b_lblkno < -maxblk-(NIADDR-1))) 538 { 539 bp->b_flags |= B_INVAL | B_VFLUSH; 540 if(lfs_iscleaner_bp(bp)) { 541 lfs_freebuf(bp); 542 } else { 543 brelse(bp); 544 } 545 ++dirty; 546 } else { 547 /* 548 * This buffer is still on its free list. 549 * So don't brelse, but wake up any sleepers. 550 */ 551 bp->b_flags &= ~B_BUSY; 552 if(bp->b_flags & B_WANTED) { 553 bp->b_flags &= ~(B_WANTED|B_AGE); 554 wakeup(bp); 555 } 556 } 557 } 558 } 559 if(dirty) 560 goto top; 561 return (0); 562 } 563
Indexes created Thu Oct 02 14:10:14 GMT 2025