lfs_inode.c revision 1.22.2.5
1 /* $NetBSD: lfs_inode.c,v 1.22.2.5 2000/01/20 21:02:03 he 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, bp) 102 struct lfs *fs; 103 ino_t ino; 104 struct buf *bp; 105 { 106 register int cnt; 107 register struct dinode *dip = (struct dinode *)bp->b_data; 108 register struct dinode *ldip; 109 110 for (cnt = INOPB(fs), ldip = dip + (cnt - 1); cnt--; --ldip) 111 if (ldip->di_inumber == ino) 112 return (ldip); 113 114 printf("offset is %d (seg %d)\n", fs->lfs_offset, datosn(fs,fs->lfs_offset)); 115 printf("block is %d (seg %d)\n", bp->b_blkno, datosn(fs,bp->b_blkno)); 116 panic("lfs_ifind: dinode %u not found", ino); 117 /* NOTREACHED */ 118 } 119 120 int 121 lfs_update(v) 122 void *v; 123 { 124 struct vop_update_args /* { 125 struct vnode *a_vp; 126 struct timespec *a_access; 127 struct timespec *a_modify; 128 int a_waitfor; 129 } */ *ap = v; 130 struct inode *ip; 131 struct vnode *vp = ap->a_vp; 132 int mod, oflag; 133 struct timespec ts; 134 struct lfs *fs = VFSTOUFS(vp->v_mount)->um_lfs; 135 136 if (vp->v_mount->mnt_flag & MNT_RDONLY) 137 return (0); 138 ip = VTOI(vp); 139 140 /* 141 * If we are called from vinvalbuf, and the file's blocks have 142 * already been scheduled for writing, but the writes have not 143 * yet completed, lfs_vflush will not be called, and vinvalbuf 144 * will cause a panic. So, we must wait until any pending write 145 * for our inode completes, if we are called with LFS_SYNC set. 146 */ 147 while((ap->a_waitfor & LFS_SYNC) && WRITEINPROG(vp)) { 148 #ifdef DEBUG_LFS 149 printf("lfs_update: sleeping on inode %d (in-progress)\n",ip->i_number); 150 #endif 151 tsleep(vp, (PRIBIO+1), "lfs_update", 0); 152 } 153 mod = ip->i_flag & IN_MODIFIED; 154 oflag = ip->i_flag; 155 TIMEVAL_TO_TIMESPEC(&time, &ts); 156 LFS_ITIMES(ip, 157 ap->a_access ? ap->a_access : &ts, 158 ap->a_modify ? ap->a_modify : &ts, &ts); 159 if (!mod && (ip->i_flag & IN_MODIFIED)) 160 ip->i_lfs->lfs_uinodes++; 161 if ((ip->i_flag & (IN_MODIFIED|IN_CLEANING)) == 0) { 162 return (0); 163 } 164 165 /* If sync, push back the vnode and any dirty blocks it may have. */ 166 if(ap->a_waitfor & LFS_SYNC) { 167 /* Avoid flushing VDIROP. */ 168 ++fs->lfs_diropwait; 169 while(vp->v_flag & VDIROP) { 170 #ifdef DEBUG_LFS 171 printf("lfs_update: sleeping on inode %d (dirops)\n",ip->i_number); 172 #endif 173 if(fs->lfs_dirops==0) 174 lfs_flush_fs(vp->v_mount,SEGM_SYNC); 175 else 176 tsleep(&fs->lfs_writer, PRIBIO+1, "lfs_fsync", 0); 177 /* XXX KS - by falling out here, are we writing the vn 178 twice? */ 179 } 180 --fs->lfs_diropwait; 181 return lfs_vflush(vp); 182 } 183 return 0; 184 } 185 186 /* Update segment usage information when removing a block. */ 187 #define UPDATE_SEGUSE \ 188 if (lastseg != -1) { \ 189 LFS_SEGENTRY(sup, fs, lastseg, sup_bp); \ 190 if (num > sup->su_nbytes) { \ 191 printf("lfs_truncate: negative bytes: segment %d short by %d\n", \ 192 lastseg, num - sup->su_nbytes); \ 193 panic("lfs_truncate: negative bytes"); \ 194 sup->su_nbytes = 0; \ 195 } else \ 196 sup->su_nbytes -= num; \ 197 e1 = VOP_BWRITE(sup_bp); \ 198 fragsreleased += numfrags(fs, num); \ 199 } 200 201 #define SEGDEC(S) { \ 202 if (daddr != 0) { \ 203 if (lastseg != (seg = datosn(fs, daddr))) { \ 204 UPDATE_SEGUSE; \ 205 num = (S); \ 206 lastseg = seg; \ 207 } else \ 208 num += (S); \ 209 } \ 210 } 211 212 /* 213 * Truncate the inode ip to at most length size. Update segment usage 214 * table information. 215 */ 216 /* ARGSUSED */ 217 int 218 lfs_truncate(v) 219 void *v; 220 { 221 struct vop_truncate_args /* { 222 struct vnode *a_vp; 223 off_t a_length; 224 int a_flags; 225 struct ucred *a_cred; 226 struct proc *a_p; 227 } */ *ap = v; 228 register struct indir *inp; 229 register int i; 230 register ufs_daddr_t *daddrp; 231 register struct vnode *vp = ap->a_vp; 232 off_t length = ap->a_length; 233 struct buf *bp, *sup_bp; 234 struct ifile *ifp; 235 struct inode *ip; 236 struct lfs *fs; 237 struct indir a[NIADDR + 2], a_end[NIADDR + 2]; 238 SEGUSE *sup; 239 ufs_daddr_t daddr, lastblock, lbn, olastblock; 240 ufs_daddr_t oldsize_lastblock, oldsize_newlast, newsize; 241 long off, a_released, fragsreleased, i_released; 242 int e1, e2, depth, lastseg, num, offset, seg, freesize; 243 244 ip = VTOI(vp); 245 246 if (vp->v_type == VLNK && vp->v_mount->mnt_maxsymlinklen > 0) { 247 #ifdef DIAGNOSTIC 248 if (length != 0) 249 panic("lfs_truncate: partial truncate of symlink"); 250 #endif 251 bzero((char *)&ip->i_ffs_shortlink, (u_int)ip->i_ffs_size); 252 ip->i_ffs_size = 0; 253 ip->i_flag |= IN_CHANGE | IN_UPDATE; 254 return (VOP_UPDATE(vp, NULL, NULL, 0)); 255 } 256 uvm_vnp_setsize(vp, length); 257 258 fs = ip->i_lfs; 259 lfs_imtime(fs); 260 261 /* If length is larger than the file, just update the times. */ 262 if (ip->i_ffs_size <= length) { 263 ip->i_flag |= IN_CHANGE | IN_UPDATE; 264 return (VOP_UPDATE(vp, NULL, NULL, 0)); 265 } 266 267 /* 268 * Make sure no writes happen while we're truncating. 269 * Otherwise, blocks which are accounted for on the inode 270 * *and* which have been created for cleaning can coexist, 271 * and cause us to overcount, and panic below. 272 * 273 * XXX KS - too restrictive? Maybe only when cleaning? 274 */ 275 while(fs->lfs_seglock && fs->lfs_lockpid != ap->a_p->p_pid) { 276 tsleep(&fs->lfs_seglock, (PRIBIO+1), "lfs_truncate", 0); 277 } 278 279 /* 280 * Calculate index into inode's block list of last direct and indirect 281 * blocks (if any) which we want to keep. Lastblock is 0 when the 282 * file is truncated to 0. 283 */ 284 lastblock = lblkno(fs, length + fs->lfs_bsize - 1); 285 olastblock = lblkno(fs, ip->i_ffs_size + fs->lfs_bsize - 1) - 1; 286 287 /* 288 * Update the size of the file. If the file is not being truncated to 289 * a block boundry, the contents of the partial block following the end 290 * must be zero'd in case it ever becomes accessible again 291 * because of subsequent file growth. For this part of the code, 292 * oldsize_newlast refers to the old size of the new last block in the 293 * file. 294 */ 295 offset = blkoff(fs, length); 296 lbn = lblkno(fs, length); 297 oldsize_newlast = blksize(fs, ip, lbn); 298 299 /* Now set oldsize to the current size of the current last block */ 300 oldsize_lastblock = blksize(fs, ip, olastblock); 301 if (offset == 0) 302 ip->i_ffs_size = length; 303 else { 304 #ifdef QUOTA 305 if ((e1 = getinoquota(ip)) != 0) 306 return (e1); 307 #endif 308 if ((e1 = bread(vp, lbn, oldsize_newlast, NOCRED, &bp)) != 0) { 309 printf("lfs_truncate: bread: %d\n",e1); 310 brelse(bp); 311 return (e1); 312 } 313 ip->i_ffs_size = length; 314 (void)uvm_vnp_uncache(vp); 315 newsize = blksize(fs, ip, lbn); 316 bzero((char *)bp->b_data + offset, (u_int)(newsize - offset)); 317 #ifdef DEBUG 318 if(bp->b_flags & B_CALL) 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 lastseg = -1; 337 338 for (lbn = olastblock; lbn >= lastblock;) { 339 /* XXX use run length from bmap array to make this faster */ 340 ufs_bmaparray(vp, lbn, &daddr, a, &depth, NULL); 341 if (lbn == olastblock) { 342 for (i = NIADDR + 2; i--;) 343 a_end[i] = a[i]; 344 freesize = oldsize_lastblock; 345 } else 346 freesize = fs->lfs_bsize; 347 switch (depth) { 348 case 0: /* Direct block. */ 349 daddr = ip->i_ffs_db[lbn]; 350 SEGDEC(freesize); 351 ip->i_ffs_db[lbn] = 0; 352 --lbn; 353 break; 354 #ifdef DIAGNOSTIC 355 case 1: /* An indirect block. */ 356 panic("lfs_truncate: ufs_bmaparray returned depth 1"); 357 /* NOTREACHED */ 358 #endif 359 default: /* Chain of indirect blocks. */ 360 inp = a + --depth; 361 if (inp->in_off > 0 && lbn != lastblock) { 362 lbn -= inp->in_off < lbn - lastblock ? 363 inp->in_off : lbn - lastblock; 364 break; 365 } 366 for (; depth && (inp->in_off == 0 || lbn == lastblock); 367 --inp, --depth) { 368 if (bread(vp, 369 inp->in_lbn, fs->lfs_bsize, NOCRED, &bp)) 370 panic("lfs_truncate: bread bno %d", 371 inp->in_lbn); 372 daddrp = (ufs_daddr_t *)bp->b_data + inp->in_off; 373 for (i = inp->in_off; 374 i++ <= a_end[depth].in_off;) { 375 daddr = *daddrp++; 376 SEGDEC(freesize); 377 } 378 a_end[depth].in_off = NINDIR(fs) - 1; 379 if (inp->in_off == 0) 380 brelse (bp); 381 else { 382 bzero((ufs_daddr_t *)bp->b_data + 383 inp->in_off, fs->lfs_bsize - 384 inp->in_off * sizeof(ufs_daddr_t)); 385 if ((e1 = VOP_BWRITE(bp)) != 0) { 386 printf("lfs_truncate: indir bwrite: %d\n",e1); 387 return (e1); 388 } 389 } 390 } 391 if (depth == 0 && a[1].in_off == 0) { 392 off = a[0].in_off; 393 daddr = ip->i_ffs_ib[off]; 394 SEGDEC(freesize); 395 ip->i_ffs_ib[off] = 0; 396 } 397 if (lbn == lastblock || lbn <= NDADDR) 398 --lbn; 399 else { 400 lbn -= NINDIR(fs); 401 if (lbn < lastblock) 402 lbn = lastblock; 403 } 404 } 405 } 406 UPDATE_SEGUSE; 407 408 /* If truncating the file to 0, update the version number. */ 409 if (length == 0) { 410 LFS_IENTRY(ifp, fs, ip->i_number, bp); 411 ++ifp->if_version; 412 (void) VOP_BWRITE(bp); 413 } 414 #ifdef DIAGNOSTIC 415 if (ip->i_ffs_blocks < fragstodb(fs, fragsreleased)) { 416 panic("lfs_truncate: frag count < 0 (%d<%ld), ino %d\n", 417 ip->i_ffs_blocks, fragstodb(fs, fragsreleased), 418 ip->i_number); 419 fragsreleased = dbtofrags(fs, ip->i_ffs_blocks); 420 } 421 #endif 422 ip->i_ffs_blocks -= fragstodb(fs, fragsreleased); 423 fs->lfs_bfree += fragstodb(fs, fragsreleased); 424 ip->i_flag |= IN_CHANGE | IN_UPDATE; 425 /* 426 * Traverse dirty block list counting number of dirty buffers 427 * that are being deleted out of the cache, so that the lfs_avail 428 * field can be updated. 429 */ 430 a_released = 0; 431 i_released = 0; 432 for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = bp->b_vnbufs.le_next) { 433 434 /* XXX KS - Don't miscount if we're not truncating to zero. */ 435 if(length>0 && !(bp->b_lblkno >= 0 && bp->b_lblkno > lastblock) 436 && !(bp->b_lblkno < 0 && bp->b_lblkno < -lastblock-NIADDR)) 437 continue; 438 439 if (bp->b_flags & B_LOCKED) { 440 a_released += numfrags(fs, bp->b_bcount); 441 /* 442 * XXX 443 * When buffers are created in the cache, their block 444 * number is set equal to their logical block number. 445 * If that is still true, we are assuming that the 446 * blocks are new (not yet on disk) and weren't 447 * counted above. However, there is a slight chance 448 * that a block's disk address is equal to its logical 449 * block number in which case, we'll get an overcounting 450 * here. 451 */ 452 if (bp->b_blkno == bp->b_lblkno) { 453 i_released += numfrags(fs, bp->b_bcount); 454 } 455 } 456 } 457 fragsreleased = i_released; 458 #ifdef DIAGNOSTIC 459 if (fragsreleased > dbtofrags(fs, ip->i_ffs_blocks)) { 460 printf("lfs_inode: %ld frags released > %d in inode %d\n", 461 fragsreleased, dbtofrags(fs, ip->i_ffs_blocks), 462 ip->i_number); 463 fragsreleased = dbtofrags(fs, ip->i_ffs_blocks); 464 } 465 #endif 466 fs->lfs_bfree += fragstodb(fs, fragsreleased); 467 ip->i_ffs_blocks -= fragstodb(fs, fragsreleased); 468 #ifdef DIAGNOSTIC 469 if (length == 0 && ip->i_ffs_blocks != 0) { 470 printf("lfs_inode: trunc to zero, but %d blocks left on inode %d\n", 471 ip->i_ffs_blocks, ip->i_number); 472 panic("lfs_inode\n"); 473 } 474 #endif 475 fs->lfs_avail += fragstodb(fs, a_released); 476 if(length>0) 477 e1 = lfs_vinvalbuf(vp, ap->a_cred, ap->a_p, lastblock-1); 478 else 479 e1 = vinvalbuf(vp, 0, ap->a_cred, ap->a_p, 0, 0); 480 e2 = VOP_UPDATE(vp, NULL, NULL, 0); 481 if(e1) 482 printf("lfs_truncate: vinvalbuf: %d\n",e1); 483 if(e2) 484 printf("lfs_truncate: update: %d\n",e2); 485 486 return (e1 ? e1 : e2 ? e2 : 0); 487 } 488 489 /* 490 * Get rid of blocks a la vinvalbuf; but only blocks that are of a higher 491 * lblkno than the file size allows. 492 */ 493 int 494 lfs_vinvalbuf(vp, cred, p, maxblk) 495 register struct vnode *vp; 496 struct ucred *cred; 497 struct proc *p; 498 ufs_daddr_t maxblk; 499 { 500 register struct buf *bp; 501 struct buf *nbp, *blist; 502 int i, s, error, dirty; 503 504 top: 505 dirty=0; 506 for (i=0;i<2;i++) { 507 if(i==0) 508 blist = vp->v_cleanblkhd.lh_first; 509 else /* i == 1 */ 510 blist = vp->v_dirtyblkhd.lh_first; 511 512 for (bp = blist; bp; bp = nbp) { 513 nbp = bp->b_vnbufs.le_next; 514 515 s = splbio(); 516 if (bp->b_flags & B_GATHERED) { 517 error = tsleep(vp, PRIBIO+1, "lfs_vin2", 0); 518 splx(s); 519 if(error) 520 return error; 521 goto top; 522 } 523 if (bp->b_flags & B_BUSY) { 524 bp->b_flags |= B_WANTED; 525 error = tsleep((caddr_t)bp, 526 (PRIBIO + 1), "lfs_vinval", 0); 527 splx(s); 528 if (error) 529 return (error); 530 goto top; 531 } 532 533 bp->b_flags |= B_BUSY; 534 splx(s); 535 if((bp->b_lblkno >= 0 && bp->b_lblkno > maxblk) 536 || (bp->b_lblkno < 0 && bp->b_lblkno < -maxblk-(NIADDR-1))) 537 { 538 bp->b_flags |= B_INVAL | B_VFLUSH; 539 if(bp->b_flags & B_CALL) { 540 lfs_freebuf(bp); 541 } else { 542 brelse(bp); 543 } 544 ++dirty; 545 } else { 546 /* 547 * This buffer is still on its free list. 548 * So don't brelse, but wake up any sleepers. 549 */ 550 bp->b_flags &= ~B_BUSY; 551 if(bp->b_flags & B_WANTED) { 552 bp->b_flags &= ~(B_WANTED|B_AGE); 553 wakeup(bp); 554 } 555 } 556 } 557 } 558 if(dirty) 559 goto top; 560 return (0); 561 } 562
Indexes created Thu Oct 02 14:10:14 GMT 2025