xref: /netbsd-src/sys/ufs/lfs/lfs_inode.c (revision 8ac07aec990b9d2e483062509d0a9fa5b4f57cf2)
1 /*	$NetBSD: lfs_inode.c,v 1.119 2008/03/27 19:06:52 ad 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@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. Neither the name of the University nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  *
66  *	@(#)lfs_inode.c	8.9 (Berkeley) 5/8/95
67  */
68 
69 #include <sys/cdefs.h>
70 __KERNEL_RCSID(0, "$NetBSD: lfs_inode.c,v 1.119 2008/03/27 19:06:52 ad Exp $");
71 
72 #if defined(_KERNEL_OPT)
73 #include "opt_quota.h"
74 #endif
75 
76 #include <sys/param.h>
77 #include <sys/systm.h>
78 #include <sys/mount.h>
79 #include <sys/malloc.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/trace.h>
86 #include <sys/resourcevar.h>
87 #include <sys/kauth.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_update_seguse(struct lfs *, struct inode *ip, long, size_t);
98 static int lfs_indirtrunc (struct inode *, daddr_t, daddr_t,
99 			   daddr_t, int, long *, long *, long *, size_t *);
100 static int lfs_blkfree (struct lfs *, struct inode *, daddr_t, size_t, long *, size_t *);
101 static int lfs_vtruncbuf(struct vnode *, daddr_t, bool, int);
102 
103 /* Search a block for a specific dinode. */
104 struct ufs1_dinode *
105 lfs_ifind(struct lfs *fs, ino_t ino, struct buf *bp)
106 {
107 	struct ufs1_dinode *dip = (struct ufs1_dinode *)bp->b_data;
108 	struct ufs1_dinode *ldip, *fin;
109 
110 	ASSERT_NO_SEGLOCK(fs);
111 	/*
112 	 * Read the inode block backwards, since later versions of the
113 	 * inode will supercede earlier ones.  Though it is unlikely, it is
114 	 * possible that the same inode will appear in the same inode block.
115 	 */
116 	fin = dip + INOPB(fs);
117 	for (ldip = fin - 1; ldip >= dip; --ldip)
118 		if (ldip->di_inumber == ino)
119 			return (ldip);
120 
121 	printf("searched %d entries\n", (int)(fin - dip));
122 	printf("offset is 0x%x (seg %d)\n", fs->lfs_offset,
123 	       dtosn(fs, fs->lfs_offset));
124 	printf("block is 0x%llx (seg %lld)\n",
125 	       (unsigned long long)dbtofsb(fs, bp->b_blkno),
126 	       (long long)dtosn(fs, dbtofsb(fs, bp->b_blkno)));
127 
128 	return NULL;
129 }
130 
131 int
132 lfs_update(struct vnode *vp, const struct timespec *acc,
133     const struct timespec *mod, int updflags)
134 {
135 	struct inode *ip;
136 	struct lfs *fs = VFSTOUFS(vp->v_mount)->um_lfs;
137 	int flags;
138 
139 	ASSERT_NO_SEGLOCK(fs);
140 	if (vp->v_mount->mnt_flag & MNT_RDONLY)
141 		return (0);
142 	ip = VTOI(vp);
143 
144 	/*
145 	 * If we are called from vinvalbuf, and the file's blocks have
146 	 * already been scheduled for writing, but the writes have not
147 	 * yet completed, lfs_vflush will not be called, and vinvalbuf
148 	 * will cause a panic.	So, we must wait until any pending write
149 	 * for our inode completes, if we are called with UPDATE_WAIT set.
150 	 */
151 	mutex_enter(&vp->v_interlock);
152 	while ((updflags & (UPDATE_WAIT|UPDATE_DIROP)) == UPDATE_WAIT &&
153 	    WRITEINPROG(vp)) {
154 		DLOG((DLOG_SEG, "lfs_update: sleeping on ino %d"
155 		      " (in progress)\n", ip->i_number));
156 		cv_wait(&vp->v_cv, &vp->v_interlock);
157 	}
158 	mutex_exit(&vp->v_interlock);
159 	LFS_ITIMES(ip, acc, mod, NULL);
160 	if (updflags & UPDATE_CLOSE)
161 		flags = ip->i_flag & (IN_MODIFIED | IN_ACCESSED | IN_CLEANING);
162 	else
163 		flags = ip->i_flag & (IN_MODIFIED | IN_CLEANING);
164 	if (flags == 0)
165 		return (0);
166 
167 	/* If sync, push back the vnode and any dirty blocks it may have. */
168 	if ((updflags & (UPDATE_WAIT|UPDATE_DIROP)) == UPDATE_WAIT) {
169 		/* Avoid flushing VU_DIROP. */
170 		mutex_enter(&lfs_lock);
171 		++fs->lfs_diropwait;
172 		while (vp->v_uflag & VU_DIROP) {
173 			DLOG((DLOG_DIROP, "lfs_update: sleeping on inode %d"
174 			      " (dirops)\n", ip->i_number));
175 			DLOG((DLOG_DIROP, "lfs_update: vflags 0x%x, iflags"
176 			      " 0x%x\n",
177 			      vp->v_iflag | vp->v_vflag | vp->v_uflag,
178 			      ip->i_flag));
179 			if (fs->lfs_dirops == 0)
180 				lfs_flush_fs(fs, SEGM_SYNC);
181 			else
182 				mtsleep(&fs->lfs_writer, PRIBIO+1, "lfs_fsync",
183 					0, &lfs_lock);
184 			/* XXX KS - by falling out here, are we writing the vn
185 			twice? */
186 		}
187 		--fs->lfs_diropwait;
188 		mutex_exit(&lfs_lock);
189 		return lfs_vflush(vp);
190 	}
191 	return 0;
192 }
193 
194 #define	SINGLE	0	/* index of single indirect block */
195 #define	DOUBLE	1	/* index of double indirect block */
196 #define	TRIPLE	2	/* index of triple indirect block */
197 /*
198  * Truncate the inode oip to at most length size, freeing the
199  * disk blocks.
200  */
201 /* VOP_BWRITE 1 + NIADDR + lfs_balloc == 2 + 2*NIADDR times */
202 
203 int
204 lfs_truncate(struct vnode *ovp, off_t length, int ioflag, kauth_cred_t cred)
205 {
206 	daddr_t lastblock;
207 	struct inode *oip = VTOI(ovp);
208 	daddr_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR];
209 	/* XXX ondisk32 */
210 	int32_t newblks[NDADDR + NIADDR];
211 	struct lfs *fs;
212 	struct buf *bp;
213 	int offset, size, level;
214 	long count, rcount, blocksreleased = 0, real_released = 0;
215 	int i, nblocks;
216 	int aflags, error, allerror = 0;
217 	off_t osize;
218 	long lastseg;
219 	size_t bc;
220 	int obufsize, odb;
221 	int usepc;
222 	struct ufsmount *ump = oip->i_ump;
223 
224 	if (ovp->v_type == VCHR || ovp->v_type == VBLK ||
225 	    ovp->v_type == VFIFO || ovp->v_type == VSOCK) {
226 		KASSERT(oip->i_size == 0);
227 		return 0;
228 	}
229 
230 	if (length < 0)
231 		return (EINVAL);
232 
233 	/*
234 	 * Just return and not update modification times.
235 	 */
236 	if (oip->i_size == length)
237 		return (0);
238 
239 	if (ovp->v_type == VLNK &&
240 	    (oip->i_size < ump->um_maxsymlinklen ||
241 	     (ump->um_maxsymlinklen == 0 &&
242 	      oip->i_ffs1_blocks == 0))) {
243 #ifdef DIAGNOSTIC
244 		if (length != 0)
245 			panic("lfs_truncate: partial truncate of symlink");
246 #endif
247 		memset((char *)SHORTLINK(oip), 0, (u_int)oip->i_size);
248 		oip->i_size = oip->i_ffs1_size = 0;
249 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
250 		return (lfs_update(ovp, NULL, NULL, 0));
251 	}
252 	if (oip->i_size == length) {
253 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
254 		return (lfs_update(ovp, NULL, NULL, 0));
255 	}
256 	fs = oip->i_lfs;
257 	lfs_imtime(fs);
258 	osize = oip->i_size;
259 	usepc = (ovp->v_type == VREG && ovp != fs->lfs_ivnode);
260 
261 	ASSERT_NO_SEGLOCK(fs);
262 	/*
263 	 * Lengthen the size of the file. We must ensure that the
264 	 * last byte of the file is allocated. Since the smallest
265 	 * value of osize is 0, length will be at least 1.
266 	 */
267 	if (osize < length) {
268 		if (length > ump->um_maxfilesize)
269 			return (EFBIG);
270 		aflags = B_CLRBUF;
271 		if (ioflag & IO_SYNC)
272 			aflags |= B_SYNC;
273 		if (usepc) {
274 			if (lblkno(fs, osize) < NDADDR &&
275 			    lblkno(fs, osize) != lblkno(fs, length) &&
276 			    blkroundup(fs, osize) != osize) {
277 				off_t eob;
278 
279 				eob = blkroundup(fs, osize);
280 				uvm_vnp_setwritesize(ovp, eob);
281 				error = ufs_balloc_range(ovp, osize,
282 				    eob - osize, cred, aflags);
283 				if (error)
284 					return error;
285 				if (ioflag & IO_SYNC) {
286 					mutex_enter(&ovp->v_interlock);
287 					VOP_PUTPAGES(ovp,
288 					    trunc_page(osize & fs->lfs_bmask),
289 					    round_page(eob),
290 					    PGO_CLEANIT | PGO_SYNCIO);
291 				}
292 			}
293 			uvm_vnp_setwritesize(ovp, length);
294 			error = ufs_balloc_range(ovp, length - 1, 1, cred,
295 						 aflags);
296 			if (error) {
297 				(void) lfs_truncate(ovp, osize,
298 						    ioflag & IO_SYNC, cred);
299 				return error;
300 			}
301 			uvm_vnp_setsize(ovp, length);
302 			oip->i_flag |= IN_CHANGE | IN_UPDATE;
303 			KASSERT(ovp->v_size == oip->i_size);
304 			oip->i_lfs_hiblk = lblkno(fs, oip->i_size + fs->lfs_bsize - 1) - 1;
305 			return (lfs_update(ovp, NULL, NULL, 0));
306 		} else {
307 			error = lfs_reserve(fs, ovp, NULL,
308 			    btofsb(fs, (NIADDR + 2) << fs->lfs_bshift));
309 			if (error)
310 				return (error);
311 			error = lfs_balloc(ovp, length - 1, 1, cred,
312 					   aflags, &bp);
313 			lfs_reserve(fs, ovp, NULL,
314 			    -btofsb(fs, (NIADDR + 2) << fs->lfs_bshift));
315 			if (error)
316 				return (error);
317 			oip->i_ffs1_size = oip->i_size = length;
318 			uvm_vnp_setsize(ovp, length);
319 			(void) VOP_BWRITE(bp);
320 			oip->i_flag |= IN_CHANGE | IN_UPDATE;
321 			oip->i_lfs_hiblk = lblkno(fs, oip->i_size + fs->lfs_bsize - 1) - 1;
322 			return (lfs_update(ovp, NULL, NULL, 0));
323 		}
324 	}
325 
326 	if ((error = lfs_reserve(fs, ovp, NULL,
327 	    btofsb(fs, (2 * NIADDR + 3) << fs->lfs_bshift))) != 0)
328 		return (error);
329 
330 	/*
331 	 * Shorten the size of the file. If the file is not being
332 	 * truncated to a block boundary, the contents of the
333 	 * partial block following the end of the file must be
334 	 * zero'ed in case it ever becomes accessible again because
335 	 * of subsequent file growth. Directories however are not
336 	 * zero'ed as they should grow back initialized to empty.
337 	 */
338 	offset = blkoff(fs, length);
339 	lastseg = -1;
340 	bc = 0;
341 
342 	if (ovp != fs->lfs_ivnode)
343 		lfs_seglock(fs, SEGM_PROT);
344 	if (offset == 0) {
345 		oip->i_size = oip->i_ffs1_size = length;
346 	} else if (!usepc) {
347 		lbn = lblkno(fs, length);
348 		aflags = B_CLRBUF;
349 		if (ioflag & IO_SYNC)
350 			aflags |= B_SYNC;
351 		error = lfs_balloc(ovp, length - 1, 1, cred, aflags, &bp);
352 		if (error) {
353 			lfs_reserve(fs, ovp, NULL,
354 			    -btofsb(fs, (2 * NIADDR + 3) << fs->lfs_bshift));
355 			goto errout;
356 		}
357 		obufsize = bp->b_bufsize;
358 		odb = btofsb(fs, bp->b_bcount);
359 		oip->i_size = oip->i_ffs1_size = length;
360 		size = blksize(fs, oip, lbn);
361 		if (ovp->v_type != VDIR)
362 			memset((char *)bp->b_data + offset, 0,
363 			       (u_int)(size - offset));
364 		allocbuf(bp, size, 1);
365 		if ((bp->b_flags & B_LOCKED) != 0 && bp->b_iodone == NULL) {
366 			mutex_enter(&lfs_lock);
367 			locked_queue_bytes -= obufsize - bp->b_bufsize;
368 			mutex_exit(&lfs_lock);
369 		}
370 		if (bp->b_oflags & BO_DELWRI)
371 			fs->lfs_avail += odb - btofsb(fs, size);
372 		(void) VOP_BWRITE(bp);
373 	} else { /* vp->v_type == VREG && length < osize && offset != 0 */
374 		/*
375 		 * When truncating a regular file down to a non-block-aligned
376 		 * size, we must zero the part of last block which is past
377 		 * the new EOF.  We must synchronously flush the zeroed pages
378 		 * to disk since the new pages will be invalidated as soon
379 		 * as we inform the VM system of the new, smaller size.
380 		 * We must do this before acquiring the GLOCK, since fetching
381 		 * the pages will acquire the GLOCK internally.
382 		 * So there is a window where another thread could see a whole
383 		 * zeroed page past EOF, but that's life.
384 		 */
385 		daddr_t xlbn;
386 		voff_t eoz;
387 
388 		aflags = ioflag & IO_SYNC ? B_SYNC : 0;
389 		error = ufs_balloc_range(ovp, length - 1, 1, cred, aflags);
390 		if (error) {
391 			lfs_reserve(fs, ovp, NULL,
392 				    -btofsb(fs, (2 * NIADDR + 3) << fs->lfs_bshift));
393 			goto errout;
394 		}
395 		xlbn = lblkno(fs, length);
396 		size = blksize(fs, oip, xlbn);
397 		eoz = MIN(lblktosize(fs, xlbn) + size, osize);
398 		uvm_vnp_zerorange(ovp, length, eoz - length);
399 		if (round_page(eoz) > round_page(length)) {
400 			mutex_enter(&ovp->v_interlock);
401 			error = VOP_PUTPAGES(ovp, round_page(length),
402 			    round_page(eoz),
403 			    PGO_CLEANIT | PGO_DEACTIVATE |
404 			    ((ioflag & IO_SYNC) ? PGO_SYNCIO : 0));
405 			if (error) {
406 				lfs_reserve(fs, ovp, NULL,
407 					    -btofsb(fs, (2 * NIADDR + 3) << fs->lfs_bshift));
408 				goto errout;
409 			}
410 		}
411 	}
412 
413 	genfs_node_wrlock(ovp);
414 
415 	oip->i_size = oip->i_ffs1_size = length;
416 	uvm_vnp_setsize(ovp, length);
417 	/*
418 	 * Calculate index into inode's block list of
419 	 * last direct and indirect blocks (if any)
420 	 * which we want to keep.  Lastblock is -1 when
421 	 * the file is truncated to 0.
422 	 */
423 	/* Avoid sign overflow - XXX assumes that off_t is a quad_t. */
424 	if (length > QUAD_MAX - fs->lfs_bsize)
425 		lastblock = lblkno(fs, QUAD_MAX - fs->lfs_bsize);
426 	else
427 		lastblock = lblkno(fs, length + fs->lfs_bsize - 1) - 1;
428 	lastiblock[SINGLE] = lastblock - NDADDR;
429 	lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
430 	lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
431 	nblocks = btofsb(fs, fs->lfs_bsize);
432 	/*
433 	 * Record changed file and block pointers before we start
434 	 * freeing blocks.  lastiblock values are also normalized to -1
435 	 * for calls to lfs_indirtrunc below.
436 	 */
437 	memcpy((void *)newblks, (void *)&oip->i_ffs1_db[0], sizeof newblks);
438 	for (level = TRIPLE; level >= SINGLE; level--)
439 		if (lastiblock[level] < 0) {
440 			newblks[NDADDR+level] = 0;
441 			lastiblock[level] = -1;
442 		}
443 	for (i = NDADDR - 1; i > lastblock; i--)
444 		newblks[i] = 0;
445 
446 	oip->i_size = oip->i_ffs1_size = osize;
447 	error = lfs_vtruncbuf(ovp, lastblock + 1, false, 0);
448 	if (error && !allerror)
449 		allerror = error;
450 
451 	/*
452 	 * Indirect blocks first.
453 	 */
454 	indir_lbn[SINGLE] = -NDADDR;
455 	indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1;
456 	indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
457 	for (level = TRIPLE; level >= SINGLE; level--) {
458 		bn = oip->i_ffs1_ib[level];
459 		if (bn != 0) {
460 			error = lfs_indirtrunc(oip, indir_lbn[level],
461 					       bn, lastiblock[level],
462 					       level, &count, &rcount,
463 					       &lastseg, &bc);
464 			if (error)
465 				allerror = error;
466 			real_released += rcount;
467 			blocksreleased += count;
468 			if (lastiblock[level] < 0) {
469 				if (oip->i_ffs1_ib[level] > 0)
470 					real_released += nblocks;
471 				blocksreleased += nblocks;
472 				oip->i_ffs1_ib[level] = 0;
473 				lfs_blkfree(fs, oip, bn, fs->lfs_bsize,
474 					    &lastseg, &bc);
475         			lfs_deregister_block(ovp, bn);
476 			}
477 		}
478 		if (lastiblock[level] >= 0)
479 			goto done;
480 	}
481 
482 	/*
483 	 * All whole direct blocks or frags.
484 	 */
485 	for (i = NDADDR - 1; i > lastblock; i--) {
486 		long bsize, obsize;
487 
488 		bn = oip->i_ffs1_db[i];
489 		if (bn == 0)
490 			continue;
491 		bsize = blksize(fs, oip, i);
492 		if (oip->i_ffs1_db[i] > 0) {
493 			/* Check for fragment size changes */
494 			obsize = oip->i_lfs_fragsize[i];
495 			real_released += btofsb(fs, obsize);
496 			oip->i_lfs_fragsize[i] = 0;
497 		} else
498 			obsize = 0;
499 		blocksreleased += btofsb(fs, bsize);
500 		oip->i_ffs1_db[i] = 0;
501 		lfs_blkfree(fs, oip, bn, obsize, &lastseg, &bc);
502         	lfs_deregister_block(ovp, bn);
503 	}
504 	if (lastblock < 0)
505 		goto done;
506 
507 	/*
508 	 * Finally, look for a change in size of the
509 	 * last direct block; release any frags.
510 	 */
511 	bn = oip->i_ffs1_db[lastblock];
512 	if (bn != 0) {
513 		long oldspace, newspace;
514 #if 0
515 		long olddspace;
516 #endif
517 
518 		/*
519 		 * Calculate amount of space we're giving
520 		 * back as old block size minus new block size.
521 		 */
522 		oldspace = blksize(fs, oip, lastblock);
523 #if 0
524 		olddspace = oip->i_lfs_fragsize[lastblock];
525 #endif
526 
527 		oip->i_size = oip->i_ffs1_size = length;
528 		newspace = blksize(fs, oip, lastblock);
529 		if (newspace == 0)
530 			panic("itrunc: newspace");
531 		if (oldspace - newspace > 0) {
532 			blocksreleased += btofsb(fs, oldspace - newspace);
533 		}
534 #if 0
535 		if (bn > 0 && olddspace - newspace > 0) {
536 			/* No segment accounting here, just vnode */
537 			real_released += btofsb(fs, olddspace - newspace);
538 		}
539 #endif
540 	}
541 
542 done:
543 	/* Finish segment accounting corrections */
544 	lfs_update_seguse(fs, oip, lastseg, bc);
545 #ifdef DIAGNOSTIC
546 	for (level = SINGLE; level <= TRIPLE; level++)
547 		if ((newblks[NDADDR + level] == 0) !=
548 		    ((oip->i_ffs1_ib[level]) == 0)) {
549 			panic("lfs itrunc1");
550 		}
551 	for (i = 0; i < NDADDR; i++)
552 		if ((newblks[i] == 0) != (oip->i_ffs1_db[i] == 0)) {
553 			panic("lfs itrunc2");
554 		}
555 	if (length == 0 &&
556 	    (!LIST_EMPTY(&ovp->v_cleanblkhd) || !LIST_EMPTY(&ovp->v_dirtyblkhd)))
557 		panic("lfs itrunc3");
558 #endif /* DIAGNOSTIC */
559 	/*
560 	 * Put back the real size.
561 	 */
562 	oip->i_size = oip->i_ffs1_size = length;
563 	oip->i_lfs_effnblks -= blocksreleased;
564 	oip->i_ffs1_blocks -= real_released;
565 	mutex_enter(&lfs_lock);
566 	fs->lfs_bfree += blocksreleased;
567 	mutex_exit(&lfs_lock);
568 #ifdef DIAGNOSTIC
569 	if (oip->i_size == 0 &&
570 	    (oip->i_ffs1_blocks != 0 || oip->i_lfs_effnblks != 0)) {
571 		printf("lfs_truncate: truncate to 0 but %d blks/%d effblks\n",
572 		       oip->i_ffs1_blocks, oip->i_lfs_effnblks);
573 		panic("lfs_truncate: persistent blocks");
574 	}
575 #endif
576 
577 	/*
578 	 * If we truncated to zero, take us off the paging queue.
579 	 */
580 	mutex_enter(&lfs_lock);
581 	if (oip->i_size == 0 && oip->i_flags & IN_PAGING) {
582 		oip->i_flags &= ~IN_PAGING;
583 		TAILQ_REMOVE(&fs->lfs_pchainhd, oip, i_lfs_pchain);
584 	}
585 	mutex_exit(&lfs_lock);
586 
587 	oip->i_flag |= IN_CHANGE;
588 #ifdef QUOTA
589 	(void) chkdq(oip, -blocksreleased, NOCRED, 0);
590 #endif
591 	lfs_reserve(fs, ovp, NULL,
592 	    -btofsb(fs, (2 * NIADDR + 3) << fs->lfs_bshift));
593 	genfs_node_unlock(ovp);
594   errout:
595 	oip->i_lfs_hiblk = lblkno(fs, oip->i_size + fs->lfs_bsize - 1) - 1;
596 	if (ovp != fs->lfs_ivnode)
597 		lfs_segunlock(fs);
598 	return (allerror ? allerror : error);
599 }
600 
601 /* Update segment and avail usage information when removing a block. */
602 static int
603 lfs_blkfree(struct lfs *fs, struct inode *ip, daddr_t daddr,
604 	    size_t bsize, long *lastseg, size_t *num)
605 {
606 	long seg;
607 	int error = 0;
608 
609 	ASSERT_SEGLOCK(fs);
610 	bsize = fragroundup(fs, bsize);
611 	if (daddr > 0) {
612 		if (*lastseg != (seg = dtosn(fs, daddr))) {
613 			error = lfs_update_seguse(fs, ip, *lastseg, *num);
614 			*num = bsize;
615 			*lastseg = seg;
616 		} else
617 			*num += bsize;
618 	}
619 
620 	return error;
621 }
622 
623 /* Finish the accounting updates for a segment. */
624 static int
625 lfs_update_seguse(struct lfs *fs, struct inode *ip, long lastseg, size_t num)
626 {
627 	struct segdelta *sd;
628 	struct vnode *vp;
629 
630 	ASSERT_SEGLOCK(fs);
631 	if (lastseg < 0 || num == 0)
632 		return 0;
633 
634 	vp = ITOV(ip);
635 	LIST_FOREACH(sd, &ip->i_lfs_segdhd, list)
636 		if (sd->segnum == lastseg)
637 			break;
638 	if (sd == NULL) {
639 		sd = malloc(sizeof(*sd), M_SEGMENT, M_WAITOK);
640 		sd->segnum = lastseg;
641 		sd->num = 0;
642 		LIST_INSERT_HEAD(&ip->i_lfs_segdhd, sd, list);
643 	}
644 	sd->num += num;
645 
646 	return 0;
647 }
648 
649 static void
650 lfs_finalize_seguse(struct lfs *fs, void *v)
651 {
652 	SEGUSE *sup;
653 	struct buf *bp;
654 	struct segdelta *sd;
655 	LIST_HEAD(, segdelta) *hd = v;
656 
657 	ASSERT_SEGLOCK(fs);
658 	while((sd = LIST_FIRST(hd)) != NULL) {
659 		LIST_REMOVE(sd, list);
660 		LFS_SEGENTRY(sup, fs, sd->segnum, bp);
661 		if (sd->num > sup->su_nbytes) {
662 			printf("lfs_finalize_seguse: segment %ld short by %ld\n",
663 				sd->segnum, (long)(sd->num - sup->su_nbytes));
664 			panic("lfs_finalize_seguse: negative bytes");
665 			sup->su_nbytes = sd->num;
666 		}
667 		sup->su_nbytes -= sd->num;
668 		LFS_WRITESEGENTRY(sup, fs, sd->segnum, bp);
669 		free(sd, M_SEGMENT);
670 	}
671 }
672 
673 /* Finish the accounting updates for a segment. */
674 void
675 lfs_finalize_ino_seguse(struct lfs *fs, struct inode *ip)
676 {
677 	ASSERT_SEGLOCK(fs);
678 	lfs_finalize_seguse(fs, &ip->i_lfs_segdhd);
679 }
680 
681 /* Finish the accounting updates for a segment. */
682 void
683 lfs_finalize_fs_seguse(struct lfs *fs)
684 {
685 	ASSERT_SEGLOCK(fs);
686 	lfs_finalize_seguse(fs, &fs->lfs_segdhd);
687 }
688 
689 /*
690  * Release blocks associated with the inode ip and stored in the indirect
691  * block bn.  Blocks are free'd in LIFO order up to (but not including)
692  * lastbn.  If level is greater than SINGLE, the block is an indirect block
693  * and recursive calls to indirtrunc must be used to cleanse other indirect
694  * blocks.
695  *
696  * NB: triple indirect blocks are untested.
697  */
698 static int
699 lfs_indirtrunc(struct inode *ip, daddr_t lbn, daddr_t dbn,
700 	       daddr_t lastbn, int level, long *countp,
701 	       long *rcountp, long *lastsegp, size_t *bcp)
702 {
703 	int i;
704 	struct buf *bp;
705 	struct lfs *fs = ip->i_lfs;
706 	int32_t *bap;	/* XXX ondisk32 */
707 	struct vnode *vp;
708 	daddr_t nb, nlbn, last;
709 	int32_t *copy = NULL;	/* XXX ondisk32 */
710 	long blkcount, rblkcount, factor;
711 	int nblocks, blocksreleased = 0, real_released = 0;
712 	int error = 0, allerror = 0;
713 
714 	ASSERT_SEGLOCK(fs);
715 	/*
716 	 * Calculate index in current block of last
717 	 * block to be kept.  -1 indicates the entire
718 	 * block so we need not calculate the index.
719 	 */
720 	factor = 1;
721 	for (i = SINGLE; i < level; i++)
722 		factor *= NINDIR(fs);
723 	last = lastbn;
724 	if (lastbn > 0)
725 		last /= factor;
726 	nblocks = btofsb(fs, fs->lfs_bsize);
727 	/*
728 	 * Get buffer of block pointers, zero those entries corresponding
729 	 * to blocks to be free'd, and update on disk copy first.  Since
730 	 * double(triple) indirect before single(double) indirect, calls
731 	 * to bmap on these blocks will fail.  However, we already have
732 	 * the on disk address, so we have to set the b_blkno field
733 	 * explicitly instead of letting bread do everything for us.
734 	 */
735 	vp = ITOV(ip);
736 	bp = getblk(vp, lbn, (int)fs->lfs_bsize, 0, 0);
737 	if (bp->b_oflags & (BO_DONE | BO_DELWRI)) {
738 		/* Braces must be here in case trace evaluates to nothing. */
739 		trace(TR_BREADHIT, pack(vp, fs->lfs_bsize), lbn);
740 	} else {
741 		trace(TR_BREADMISS, pack(vp, fs->lfs_bsize), lbn);
742 		curlwp->l_ru.ru_inblock++; /* pay for read */
743 		bp->b_flags |= B_READ;
744 		if (bp->b_bcount > bp->b_bufsize)
745 			panic("lfs_indirtrunc: bad buffer size");
746 		bp->b_blkno = fsbtodb(fs, dbn);
747 		VOP_STRATEGY(vp, bp);
748 		error = biowait(bp);
749 	}
750 	if (error) {
751 		brelse(bp, 0);
752 		*countp = *rcountp = 0;
753 		return (error);
754 	}
755 
756 	bap = (int32_t *)bp->b_data;	/* XXX ondisk32 */
757 	if (lastbn >= 0) {
758 		copy = (int32_t *)lfs_malloc(fs, fs->lfs_bsize, LFS_NB_IBLOCK);
759 		memcpy((void *)copy, (void *)bap, (u_int)fs->lfs_bsize);
760 		memset((void *)&bap[last + 1], 0,
761 		/* XXX ondisk32 */
762 		  (u_int)(NINDIR(fs) - (last + 1)) * sizeof (int32_t));
763 		error = VOP_BWRITE(bp);
764 		if (error)
765 			allerror = error;
766 		bap = copy;
767 	}
768 
769 	/*
770 	 * Recursively free totally unused blocks.
771 	 */
772 	for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last;
773 	    i--, nlbn += factor) {
774 		nb = bap[i];
775 		if (nb == 0)
776 			continue;
777 		if (level > SINGLE) {
778 			error = lfs_indirtrunc(ip, nlbn, nb,
779 					       (daddr_t)-1, level - 1,
780 					       &blkcount, &rblkcount,
781 					       lastsegp, bcp);
782 			if (error)
783 				allerror = error;
784 			blocksreleased += blkcount;
785 			real_released += rblkcount;
786 		}
787 		lfs_blkfree(fs, ip, nb, fs->lfs_bsize, lastsegp, bcp);
788 		if (bap[i] > 0)
789 			real_released += nblocks;
790 		blocksreleased += nblocks;
791 	}
792 
793 	/*
794 	 * Recursively free last partial block.
795 	 */
796 	if (level > SINGLE && lastbn >= 0) {
797 		last = lastbn % factor;
798 		nb = bap[i];
799 		if (nb != 0) {
800 			error = lfs_indirtrunc(ip, nlbn, nb,
801 					       last, level - 1, &blkcount,
802 					       &rblkcount, lastsegp, bcp);
803 			if (error)
804 				allerror = error;
805 			real_released += rblkcount;
806 			blocksreleased += blkcount;
807 		}
808 	}
809 
810 	if (copy != NULL) {
811 		lfs_free(fs, copy, LFS_NB_IBLOCK);
812 	} else {
813 		mutex_enter(&bufcache_lock);
814 		if (bp->b_oflags & BO_DELWRI) {
815 			LFS_UNLOCK_BUF(bp);
816 			fs->lfs_avail += btofsb(fs, bp->b_bcount);
817 			wakeup(&fs->lfs_avail);
818 		}
819 		brelsel(bp, BC_INVAL);
820 		mutex_exit(&bufcache_lock);
821 	}
822 
823 	*countp = blocksreleased;
824 	*rcountp = real_released;
825 	return (allerror);
826 }
827 
828 /*
829  * Destroy any in core blocks past the truncation length.
830  * Inlined from vtruncbuf, so that lfs_avail could be updated.
831  * We take the seglock to prevent cleaning from occurring while we are
832  * invalidating blocks.
833  */
834 static int
835 lfs_vtruncbuf(struct vnode *vp, daddr_t lbn, bool catch, int slptimeo)
836 {
837 	struct buf *bp, *nbp;
838 	int error;
839 	struct lfs *fs;
840 	voff_t off;
841 
842 	off = round_page((voff_t)lbn << vp->v_mount->mnt_fs_bshift);
843 	mutex_enter(&vp->v_interlock);
844 	error = VOP_PUTPAGES(vp, off, 0, PGO_FREE | PGO_SYNCIO);
845 	if (error)
846 		return error;
847 
848 	fs = VTOI(vp)->i_lfs;
849 
850 	ASSERT_SEGLOCK(fs);
851 
852 	mutex_enter(&bufcache_lock);
853 restart:
854 	for (bp = LIST_FIRST(&vp->v_cleanblkhd); bp; bp = nbp) {
855 		nbp = LIST_NEXT(bp, b_vnbufs);
856 		if (bp->b_lblkno < lbn)
857 			continue;
858 		error = bbusy(bp, catch, slptimeo, NULL);
859 		if (error == EPASSTHROUGH)
860 			goto restart;
861 		if (error != 0) {
862 			mutex_exit(&bufcache_lock);
863 			return (error);
864 		}
865 		mutex_enter(bp->b_objlock);
866 		if (bp->b_oflags & BO_DELWRI) {
867 			bp->b_oflags &= ~BO_DELWRI;
868 			fs->lfs_avail += btofsb(fs, bp->b_bcount);
869 			wakeup(&fs->lfs_avail);
870 		}
871 		mutex_exit(bp->b_objlock);
872 		LFS_UNLOCK_BUF(bp);
873 		brelsel(bp, BC_INVAL | BC_VFLUSH);
874 	}
875 
876 	for (bp = LIST_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
877 		nbp = LIST_NEXT(bp, b_vnbufs);
878 		if (bp->b_lblkno < lbn)
879 			continue;
880 		error = bbusy(bp, catch, slptimeo, NULL);
881 		if (error == EPASSTHROUGH)
882 			goto restart;
883 		if (error != 0) {
884 			mutex_exit(&bufcache_lock);
885 			return (error);
886 		}
887 		mutex_enter(bp->b_objlock);
888 		if (bp->b_oflags & BO_DELWRI) {
889 			bp->b_oflags &= ~BO_DELWRI;
890 			fs->lfs_avail += btofsb(fs, bp->b_bcount);
891 			wakeup(&fs->lfs_avail);
892 		}
893 		mutex_exit(bp->b_objlock);
894 		LFS_UNLOCK_BUF(bp);
895 		brelsel(bp, BC_INVAL | BC_VFLUSH);
896 	}
897 	mutex_exit(&bufcache_lock);
898 
899 	return (0);
900 }
901 
902