Home | History | Annotate | Line # | Download | only in ext2fs
ext2fs_inode.c revision 1.11
      1 /*	$NetBSD: ext2fs_inode.c,v 1.11 1999/03/05 20:47:07 mycroft Exp $	*/
      2 
      3 /*
      4  * Copyright (c) 1997 Manuel Bouyer.
      5  * Copyright (c) 1982, 1986, 1989, 1993
      6  *	The Regents of the University of California.  All rights reserved.
      7  *
      8  * Redistribution and use in source and binary forms, with or without
      9  * modification, are permitted provided that the following conditions
     10  * are met:
     11  * 1. Redistributions of source code must retain the above copyright
     12  *    notice, this list of conditions and the following disclaimer.
     13  * 2. Redistributions in binary form must reproduce the above copyright
     14  *    notice, this list of conditions and the following disclaimer in the
     15  *    documentation and/or other materials provided with the distribution.
     16  * 3. All advertising materials mentioning features or use of this software
     17  *    must display the following acknowledgement:
     18  *	This product includes software developed by the University of
     19  *	California, Berkeley and its contributors.
     20  * 4. Neither the name of the University nor the names of its contributors
     21  *    may be used to endorse or promote products derived from this software
     22  *    without specific prior written permission.
     23  *
     24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     34  * SUCH DAMAGE.
     35  *
     36  *	@(#)ffs_inode.c	8.8 (Berkeley) 10/19/94
     37  * Modified for ext2fs by Manuel Bouyer.
     38  */
     39 
     40 #if defined(_KERNEL) && !defined(_LKM)
     41 #include "opt_uvm.h"
     42 #endif
     43 
     44 #include <sys/param.h>
     45 #include <sys/systm.h>
     46 #include <sys/mount.h>
     47 #include <sys/proc.h>
     48 #include <sys/file.h>
     49 #include <sys/buf.h>
     50 #include <sys/vnode.h>
     51 #include <sys/kernel.h>
     52 #include <sys/malloc.h>
     53 #include <sys/trace.h>
     54 #include <sys/resourcevar.h>
     55 
     56 #include <vm/vm.h>
     57 #if defined(UVM)
     58 #include <uvm/uvm_extern.h>
     59 #endif
     60 
     61 #include <ufs/ufs/quota.h>
     62 #include <ufs/ufs/inode.h>
     63 #include <ufs/ufs/ufsmount.h>
     64 #include <ufs/ufs/ufs_extern.h>
     65 
     66 #include <ufs/ext2fs/ext2fs.h>
     67 #include <ufs/ext2fs/ext2fs_extern.h>
     68 
     69 static int ext2fs_indirtrunc __P((struct inode *, ufs_daddr_t, ufs_daddr_t,
     70 				  ufs_daddr_t, int, long *));
     71 
     72 /*
     73  * Last reference to an inode.  If necessary, write or delete it.
     74  */
     75 int
     76 ext2fs_inactive(v)
     77 	void *v;
     78 {
     79 	struct vop_inactive_args /* {
     80 		struct vnode *a_vp;
     81 		struct proc *a_p;
     82 	} */ *ap = v;
     83 	struct vnode *vp = ap->a_vp;
     84 	struct inode *ip = VTOI(vp);
     85 	struct proc *p = ap->a_p;
     86 	struct timespec ts;
     87 	int error = 0;
     88 	extern int prtactive;
     89 
     90 	if (prtactive && vp->v_usecount != 0)
     91 		vprint("ext2fs_inactive: pushing active", vp);
     92 	/* Get rid of inodes related to stale file handles. */
     93 	if (ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime != 0)
     94 		goto out;
     95 
     96 	error = 0;
     97 	if (ip->i_e2fs_nlink == 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
     98 		error = VOP_TRUNCATE(vp, (off_t)0, 0, NOCRED, NULL);
     99 		TIMEVAL_TO_TIMESPEC(&time, &ts);
    100 		ip->i_e2fs_dtime = ts.tv_sec;
    101 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
    102 		VOP_VFREE(vp, ip->i_number, ip->i_e2fs_mode);
    103 	}
    104 	if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) {
    105 		TIMEVAL_TO_TIMESPEC(&time, &ts);
    106 		VOP_UPDATE(vp, &ts, &ts, 0);
    107 	}
    108 out:
    109 	VOP_UNLOCK(vp, 0);
    110 	/*
    111 	 * If we are done with the inode, reclaim it
    112 	 * so that it can be reused immediately.
    113 	 */
    114 	if (ip->i_e2fs_dtime != 0)
    115 		vrecycle(vp, (struct simplelock *)0, p);
    116 	return (error);
    117 }
    118 
    119 
    120 /*
    121  * Update the access, modified, and inode change times as specified by the
    122  * IACCESS, IUPDATE, and ICHANGE flags respectively. The IMODIFIED flag is
    123  * used to specify that the inode needs to be updated but that the times have
    124  * already been set. The access and modified times are taken from the second
    125  * and third parameters; the inode change time is always taken from the current
    126  * time. If waitfor is set, then wait for the disk write of the inode to
    127  * complete.
    128  */
    129 int
    130 ext2fs_update(v)
    131 	void *v;
    132 {
    133 	struct vop_update_args /* {
    134 		struct vnode *a_vp;
    135 		struct timespec *a_access;
    136 		struct timespec *a_modify;
    137 		int a_waitfor;
    138 	} */ *ap = v;
    139 	register struct m_ext2fs *fs;
    140 	struct buf *bp;
    141 	struct inode *ip;
    142 	int error;
    143 	struct timespec ts;
    144 	caddr_t cp;
    145 
    146 	if (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY)
    147 		return (0);
    148 	ip = VTOI(ap->a_vp);
    149 	TIMEVAL_TO_TIMESPEC(&time, &ts);
    150 	EXT2FS_ITIMES(ip,
    151 	    ap->a_access ? ap->a_access : &ts,
    152 	    ap->a_modify ? ap->a_modify : &ts, &ts);
    153 	if ((ip->i_flag & IN_MODIFIED) == 0)
    154 		return (0);
    155 	ip->i_flag &= ~IN_MODIFIED;
    156 	fs = ip->i_e2fs;
    157 	error = bread(ip->i_devvp,
    158 			  fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
    159 			  (int)fs->e2fs_bsize, NOCRED, &bp);
    160 	if (error) {
    161 		brelse(bp);
    162 		return (error);
    163 	}
    164 	cp = (caddr_t)bp->b_data +
    165 	    (ino_to_fsbo(fs, ip->i_number) * EXT2_DINODE_SIZE);
    166 	e2fs_isave(&ip->i_din.e2fs_din, (struct ext2fs_dinode *)cp);
    167 	if (ap->a_waitfor && (ap->a_vp->v_mount->mnt_flag & MNT_ASYNC) == 0)
    168 		return (bwrite(bp));
    169 	else {
    170 		bdwrite(bp);
    171 		return (0);
    172 	}
    173 }
    174 
    175 #define	SINGLE	0	/* index of single indirect block */
    176 #define	DOUBLE	1	/* index of double indirect block */
    177 #define	TRIPLE	2	/* index of triple indirect block */
    178 /*
    179  * Truncate the inode oip to at most length size, freeing the
    180  * disk blocks.
    181  */
    182 int
    183 ext2fs_truncate(v)
    184 	void *v;
    185 {
    186 	struct vop_truncate_args /* {
    187 		struct vnode *a_vp;
    188 		off_t a_length;
    189 		int a_flags;
    190 		struct ucred *a_cred;
    191 		struct proc *a_p;
    192 	} */ *ap = v;
    193 	register struct vnode *ovp = ap->a_vp;
    194 	register ufs_daddr_t lastblock;
    195 	register struct inode *oip;
    196 	ufs_daddr_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR];
    197 	ufs_daddr_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
    198 	off_t length = ap->a_length;
    199 	register struct m_ext2fs *fs;
    200 	struct buf *bp;
    201 	int offset, size, level;
    202 	long count, nblocks, vflags, blocksreleased = 0;
    203 	struct timespec ts;
    204 	register int i;
    205 	int aflags, error, allerror;
    206 	off_t osize;
    207 
    208 	if (length < 0)
    209 		return (EINVAL);
    210 
    211 	oip = VTOI(ovp);
    212 	TIMEVAL_TO_TIMESPEC(&time, &ts);
    213 	if (ovp->v_type == VLNK &&
    214 		(oip->i_e2fs_size < ovp->v_mount->mnt_maxsymlinklen ||
    215 		 (ovp->v_mount->mnt_maxsymlinklen == 0 &&
    216 		  oip->i_e2fs_nblock == 0))) {
    217 #ifdef DIAGNOSTIC
    218 		if (length != 0)
    219 			panic("ext2fs_truncate: partial truncate of symlink");
    220 #endif
    221 		memset((char *)&oip->i_din.e2fs_din.e2di_shortlink, 0,
    222 			(u_int)oip->i_e2fs_size);
    223 		oip->i_e2fs_size = 0;
    224 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
    225 		return (VOP_UPDATE(ovp, &ts, &ts, 1));
    226 	}
    227 	if (oip->i_e2fs_size == length) {
    228 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
    229 		return (VOP_UPDATE(ovp, &ts, &ts, 0));
    230 	}
    231 	fs = oip->i_e2fs;
    232 	osize = oip->i_e2fs_size;
    233 	/*
    234 	 * Lengthen the size of the file. We must ensure that the
    235 	 * last byte of the file is allocated. Since the smallest
    236 	 * value of osize is 0, length will be at least 1.
    237 	 */
    238 	if (osize < length) {
    239 #if 0 /* XXX */
    240 		if (length > fs->fs_maxfilesize)
    241 			return (EFBIG);
    242 #endif
    243 		offset = blkoff(fs, length - 1);
    244 		lbn = lblkno(fs, length - 1);
    245 		aflags = B_CLRBUF;
    246 		if (ap->a_flags & IO_SYNC)
    247 			aflags |= B_SYNC;
    248 		error = ext2fs_balloc(oip, lbn, offset + 1, ap->a_cred, &bp,
    249 				   aflags);
    250 		if (error)
    251 			return (error);
    252 		oip->i_e2fs_size = length;
    253 #if defined(UVM)
    254 		uvm_vnp_setsize(ovp, length);
    255 		(void) uvm_vnp_uncache(ovp);
    256 #else
    257 		vnode_pager_setsize(ovp, length);
    258 		(void) vnode_pager_uncache(ovp);
    259 #endif
    260 		if (aflags & B_SYNC)
    261 			bwrite(bp);
    262 		else
    263 			bawrite(bp);
    264 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
    265 		return (VOP_UPDATE(ovp, &ts, &ts, 1));
    266 	}
    267 	/*
    268 	 * Shorten the size of the file. If the file is not being
    269 	 * truncated to a block boundry, the contents of the
    270 	 * partial block following the end of the file must be
    271 	 * zero'ed in case it ever become accessable again because
    272 	 * of subsequent file growth.
    273 	 */
    274 	offset = blkoff(fs, length);
    275 	if (offset == 0) {
    276 		oip->i_e2fs_size = length;
    277 	} else {
    278 		lbn = lblkno(fs, length);
    279 		aflags = B_CLRBUF;
    280 		if (ap->a_flags & IO_SYNC)
    281 			aflags |= B_SYNC;
    282 		error = ext2fs_balloc(oip, lbn, offset, ap->a_cred, &bp, aflags);
    283 		if (error)
    284 			return (error);
    285 		oip->i_e2fs_size = length;
    286 		size = fs->e2fs_bsize;
    287 #if defined(UVM)
    288 		(void) uvm_vnp_uncache(ovp);
    289 #else
    290 		(void) vnode_pager_uncache(ovp);
    291 #endif
    292 		memset((char *)bp->b_data + offset, 0,  (u_int)(size - offset));
    293 		allocbuf(bp, size);
    294 		if (aflags & B_SYNC)
    295 			bwrite(bp);
    296 		else
    297 			bawrite(bp);
    298 	}
    299 #if defined(UVM)
    300 	uvm_vnp_setsize(ovp, length);
    301 #else
    302 	vnode_pager_setsize(ovp, length);
    303 #endif
    304 	/*
    305 	 * Calculate index into inode's block list of
    306 	 * last direct and indirect blocks (if any)
    307 	 * which we want to keep.  Lastblock is -1 when
    308 	 * the file is truncated to 0.
    309 	 */
    310 	lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1;
    311 	lastiblock[SINGLE] = lastblock - NDADDR;
    312 	lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
    313 	lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
    314 	nblocks = btodb(fs->e2fs_bsize);
    315 	/*
    316 	 * Update file and block pointers on disk before we start freeing
    317 	 * blocks.  If we crash before free'ing blocks below, the blocks
    318 	 * will be returned to the free list.  lastiblock values are also
    319 	 * normalized to -1 for calls to ext2fs_indirtrunc below.
    320 	 */
    321 	memcpy((caddr_t)oldblks, (caddr_t)&oip->i_e2fs_blocks[0], sizeof oldblks);
    322 	for (level = TRIPLE; level >= SINGLE; level--)
    323 		if (lastiblock[level] < 0) {
    324 			oip->i_e2fs_blocks[NDADDR + level] = 0;
    325 			lastiblock[level] = -1;
    326 		}
    327 	for (i = NDADDR - 1; i > lastblock; i--)
    328 		oip->i_e2fs_blocks[i] = 0;
    329 	oip->i_flag |= IN_CHANGE | IN_UPDATE;
    330 	if ((error = VOP_UPDATE(ovp, &ts, &ts, 1)) != 0)
    331 		allerror = error;
    332 	/*
    333 	 * Having written the new inode to disk, save its new configuration
    334 	 * and put back the old block pointers long enough to process them.
    335 	 * Note that we save the new block configuration so we can check it
    336 	 * when we are done.
    337 	 */
    338 	memcpy((caddr_t)newblks, (caddr_t)&oip->i_e2fs_blocks[0], sizeof newblks);
    339 	memcpy((caddr_t)&oip->i_e2fs_blocks[0], (caddr_t)oldblks, sizeof oldblks);
    340 	oip->i_e2fs_size = osize;
    341 	vflags = ((length > 0) ? V_SAVE : 0) | V_SAVEMETA;
    342 	allerror = vinvalbuf(ovp, vflags, ap->a_cred, ap->a_p, 0, 0);
    343 
    344 	/*
    345 	 * Indirect blocks first.
    346 	 */
    347 	indir_lbn[SINGLE] = -NDADDR;
    348 	indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) -1;
    349 	indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
    350 	for (level = TRIPLE; level >= SINGLE; level--) {
    351 		bn = fs2h32(oip->i_e2fs_blocks[NDADDR + level]);
    352 		if (bn != 0) {
    353 			error = ext2fs_indirtrunc(oip, indir_lbn[level],
    354 				fsbtodb(fs, bn), lastiblock[level], level, &count);
    355 			if (error)
    356 				allerror = error;
    357 			blocksreleased += count;
    358 			if (lastiblock[level] < 0) {
    359 				oip->i_e2fs_blocks[NDADDR + level] = 0;
    360 				ext2fs_blkfree(oip, bn);
    361 				blocksreleased += nblocks;
    362 			}
    363 		}
    364 		if (lastiblock[level] >= 0)
    365 			goto done;
    366 	}
    367 
    368 	/*
    369 	 * All whole direct blocks or frags.
    370 	 */
    371 	for (i = NDADDR - 1; i > lastblock; i--) {
    372 		bn = fs2h32(oip->i_e2fs_blocks[i]);
    373 		if (bn == 0)
    374 			continue;
    375 		oip->i_e2fs_blocks[i] = 0;
    376 		ext2fs_blkfree(oip, bn);
    377 		blocksreleased += btodb(fs->e2fs_bsize);
    378 	}
    379 	if (lastblock < 0)
    380 		goto done;
    381 
    382 done:
    383 #ifdef DIAGNOSTIC
    384 	for (level = SINGLE; level <= TRIPLE; level++)
    385 		if (newblks[NDADDR + level] !=
    386 			fs2h32(oip->i_e2fs_blocks[NDADDR + level]))
    387 			panic("itrunc1");
    388 	for (i = 0; i < NDADDR; i++)
    389 		if (newblks[i] != fs2h32(oip->i_e2fs_blocks[i]))
    390 			panic("itrunc2");
    391 	if (length == 0 &&
    392 		(ovp->v_dirtyblkhd.lh_first || ovp->v_cleanblkhd.lh_first))
    393 		panic("itrunc3");
    394 #endif /* DIAGNOSTIC */
    395 	/*
    396 	 * Put back the real size.
    397 	 */
    398 	oip->i_e2fs_size = length;
    399 	oip->i_e2fs_nblock -= blocksreleased;
    400 	if (oip->i_e2fs_nblock < 0)			/* sanity */
    401 		oip->i_e2fs_nblock = 0;
    402 	oip->i_flag |= IN_CHANGE;
    403 	return (allerror);
    404 }
    405 
    406 /*
    407  * Release blocks associated with the inode ip and stored in the indirect
    408  * block bn.  Blocks are free'd in LIFO order up to (but not including)
    409  * lastbn.  If level is greater than SINGLE, the block is an indirect block
    410  * and recursive calls to indirtrunc must be used to cleanse other indirect
    411  * blocks.
    412  *
    413  * NB: triple indirect blocks are untested.
    414  */
    415 static int
    416 ext2fs_indirtrunc(ip, lbn, dbn, lastbn, level, countp)
    417 	register struct inode *ip;
    418 	ufs_daddr_t lbn, lastbn;
    419 	ufs_daddr_t dbn;
    420 	int level;
    421 	long *countp;
    422 {
    423 	register int i;
    424 	struct buf *bp;
    425 	register struct m_ext2fs *fs = ip->i_e2fs;
    426 	register ufs_daddr_t *bap;
    427 	struct vnode *vp;
    428 	ufs_daddr_t *copy = NULL, nb, nlbn, last;
    429 	long blkcount, factor;
    430 	int nblocks, blocksreleased = 0;
    431 	int error = 0, allerror = 0;
    432 
    433 	/*
    434 	 * Calculate index in current block of last
    435 	 * block to be kept.  -1 indicates the entire
    436 	 * block so we need not calculate the index.
    437 	 */
    438 	factor = 1;
    439 	for (i = SINGLE; i < level; i++)
    440 		factor *= NINDIR(fs);
    441 	last = lastbn;
    442 	if (lastbn > 0)
    443 		last /= factor;
    444 	nblocks = btodb(fs->e2fs_bsize);
    445 	/*
    446 	 * Get buffer of block pointers, zero those entries corresponding
    447 	 * to blocks to be free'd, and update on disk copy first.  Since
    448 	 * double(triple) indirect before single(double) indirect, calls
    449 	 * to bmap on these blocks will fail.  However, we already have
    450 	 * the on disk address, so we have to set the b_blkno field
    451 	 * explicitly instead of letting bread do everything for us.
    452 	 */
    453 	vp = ITOV(ip);
    454 	bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0);
    455 	if (bp->b_flags & (B_DONE | B_DELWRI)) {
    456 		/* Braces must be here in case trace evaluates to nothing. */
    457 		trace(TR_BREADHIT, pack(vp, fs->e2fs_bsize), lbn);
    458 	} else {
    459 		trace(TR_BREADMISS, pack(vp, fs->e2fs_bsize), lbn);
    460 		curproc->p_stats->p_ru.ru_inblock++;	/* pay for read */
    461 		bp->b_flags |= B_READ;
    462 		if (bp->b_bcount > bp->b_bufsize)
    463 			panic("ext2fs_indirtrunc: bad buffer size");
    464 		bp->b_blkno = dbn;
    465 		VOP_STRATEGY(bp);
    466 		error = biowait(bp);
    467 	}
    468 	if (error) {
    469 		brelse(bp);
    470 		*countp = 0;
    471 		return (error);
    472 	}
    473 
    474 	bap = (ufs_daddr_t *)bp->b_data;
    475 	if (lastbn != -1) {
    476 		MALLOC(copy, ufs_daddr_t *, fs->e2fs_bsize, M_TEMP, M_WAITOK);
    477 		memcpy((caddr_t)copy, (caddr_t)bap, (u_int)fs->e2fs_bsize);
    478 		memset((caddr_t)&bap[last + 1], 0,
    479 			(u_int)(NINDIR(fs) - (last + 1)) * sizeof (u_int32_t));
    480 		error = bwrite(bp);
    481 		if (error)
    482 			allerror = error;
    483 		bap = copy;
    484 	}
    485 
    486 	/*
    487 	 * Recursively free totally unused blocks.
    488 	 */
    489 	for (i = NINDIR(fs) - 1,
    490 		nlbn = lbn + 1 - i * factor; i > last;
    491 		i--, nlbn += factor) {
    492 		nb = fs2h32(bap[i]);
    493 		if (nb == 0)
    494 			continue;
    495 		if (level > SINGLE) {
    496 			error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
    497 						   (ufs_daddr_t)-1, level - 1,
    498 						   &blkcount);
    499 			if (error)
    500 				allerror = error;
    501 			blocksreleased += blkcount;
    502 		}
    503 		ext2fs_blkfree(ip, nb);
    504 		blocksreleased += nblocks;
    505 	}
    506 
    507 	/*
    508 	 * Recursively free last partial block.
    509 	 */
    510 	if (level > SINGLE && lastbn >= 0) {
    511 		last = lastbn % factor;
    512 		nb = fs2h32(bap[i]);
    513 		if (nb != 0) {
    514 			error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
    515 						   last, level - 1, &blkcount);
    516 			if (error)
    517 				allerror = error;
    518 			blocksreleased += blkcount;
    519 		}
    520 	}
    521 
    522 	if (copy != NULL) {
    523 		FREE(copy, M_TEMP);
    524 	} else {
    525 		bp->b_flags |= B_INVAL;
    526 		brelse(bp);
    527 	}
    528 
    529 	*countp = blocksreleased;
    530 	return (allerror);
    531 }
    532