root/ufs/ext2fs/ext2fs_inode.c

DEFINITIONS

1. ext2fs_size
2. ext2fs_setsize
3. ext2fs_inactive
4. ext2fs_update
5. ext2fs_truncate
6. ext2fs_indirtrunc

    1 /*      $OpenBSD: ext2fs_inode.c,v 1.38 2007/06/17 20:15:25 jasper Exp $        */
    2 /*      $NetBSD: ext2fs_inode.c,v 1.24 2001/06/19 12:59:18 wiz Exp $    */
    4 /*
    5  * Copyright (c) 1997 Manuel Bouyer.
    6  * Copyright (c) 1982, 1986, 1989, 1993
    7  *      The Regents of the University of California.  All rights reserved.
    8  *
    9  * Redistribution and use in source and binary forms, with or without
   10  * modification, are permitted provided that the following conditions
   11  * are met:
   12  * 1. Redistributions of source code must retain the above copyright
   13  *    notice, this list of conditions and the following disclaimer.
   14  * 2. Redistributions in binary form must reproduce the above copyright
   15  *    notice, this list of conditions and the following disclaimer in the
   16  *    documentation and/or other materials provided with the distribution.
   17  * 3. Neither the name of the University nor the names of its contributors
   18  *    may be used to endorse or promote products derived from this software
   19  *    without specific prior written permission.
   20  *
   21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
   22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
   25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   31  * SUCH DAMAGE.
   32  *
   33  *      @(#)ffs_inode.c 8.8 (Berkeley) 10/19/94
   34  * Modified for ext2fs by Manuel Bouyer.
   35  */
   37 #include <sys/param.h>
   38 #include <sys/systm.h>
   39 #include <sys/mount.h>
   40 #include <sys/proc.h>
   41 #include <sys/file.h>
   42 #include <sys/buf.h>
   43 #include <sys/vnode.h>
   44 #include <sys/kernel.h>
   45 #include <sys/malloc.h>
   46 #include <sys/resourcevar.h>
   48 #include <uvm/uvm_extern.h>
   50 #include <ufs/ufs/quota.h>
   51 #include <ufs/ufs/inode.h>
   52 #include <ufs/ufs/ufsmount.h>
   53 #include <ufs/ufs/ufs_extern.h>
   55 #include <ufs/ext2fs/ext2fs.h>
   56 #include <ufs/ext2fs/ext2fs_extern.h>
   58 static int ext2fs_indirtrunc(struct inode *, int32_t, int32_t,
int32_t, int, long *);
   61 /*
   62  * Get the size of an inode.
   63  */
u_int64_t
ext2fs_size(struct inode *ip)
   66 {
u_int64_t size = ip->i_e2fs_size;
   69         if ((ip->i_e2fs_mode & IFMT) == IFREG)
size |= (u_int64_t)ip->i_e2fs_dacl << 32;
   72         return (size);
   73 }
   75 int
ext2fs_setsize(struct inode *ip, u_int64_t size)
   77 {
   78         if ((ip->i_e2fs_mode & IFMT) == IFREG ||
ip->i_e2fs_mode == 0) {
ip->i_e2fs_dacl = size >> 32;
   81                 if (size >= 0x80000000U) {
   82                         struct m_ext2fs *fs = ip->i_e2fs;
   84                         if (fs->e2fs.e2fs_rev <= E2FS_REV0) {
   85                                 /* Linux automagically upgrades to REV1 here! */
   86                                 return (EFBIG);
   87                         }
   88                         if (!(fs->e2fs.e2fs_features_rocompat
   89                             & EXT2F_ROCOMPAT_LARGEFILE)) {
fs->e2fs.e2fs_features_rocompat |=
EXT2F_ROCOMPAT_LARGEFILE;
fs->e2fs_fmod = 1;
   93                         }
   94                 }
   95         } else if (size >= 0x80000000U)
   96                 return (EFBIG);
ip->i_e2fs_size = size;
  100         return (0);
  101 }
  104 /*
  105  * Last reference to an inode.  If necessary, write or delete it.
  106  */
  107 int
ext2fs_inactive(void *v)
  109 {   
  110         struct vop_inactive_args *ap = v;
  111         struct vnode *vp = ap->a_vp;
  112         struct inode *ip = VTOI(vp);
  113         struct proc *p = ap->a_p;
  114         struct timespec ts;
  115         int error = 0;
  116 #ifdef DIAGNOSTIC
  117         extern int prtactive;
  119         if (prtactive && vp->v_usecount != 0)
vprint("ext2fs_inactive: pushing active", vp);
  121 #endif
  123         /* Get rid of inodes related to stale file handles. */
  124         if (ip->i_e2din == NULL || ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime)
  125                 goto out;
error = 0;
  128         if (ip->i_e2fs_nlink == 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
  129                 if (ext2fs_size(ip) != 0) {
error = ext2fs_truncate(ip, (off_t)0, 0, NOCRED);
  131                 }
getnanotime(&ts);
ip->i_e2fs_dtime = ts.tv_sec;
ip->i_flag |= IN_CHANGE | IN_UPDATE;
ext2fs_inode_free(ip, ip->i_number, ip->i_e2fs_mode);
  136         }
  137         if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) {
ext2fs_update(ip, NULL, NULL, 0);
  139         }
out:
VOP_UNLOCK(vp, 0, p);
  142         /*
  143          * If we are done with the inode, reclaim it
  144          * so that it can be reused immediately.
  145          */
  146         if (ip->i_e2din == NULL || ip->i_e2fs_dtime != 0)
vrecycle(vp, p);
  148         return (error);
  149 }   
  152 /*
  153  * Update the access, modified, and inode change times as specified by the
  154  * IACCESS, IUPDATE, and ICHANGE flags respectively. The IMODIFIED flag is
  155  * used to specify that the inode needs to be updated but that the times have
  156  * already been set. The access and modified times are taken from the second
  157  * and third parameters; the inode change time is always taken from the current
  158  * time. If waitfor is set, then wait for the disk write of the inode to
  159  * complete.
  160  */
  161 int
ext2fs_update(struct inode *ip, struct timespec *atime, struct timespec *mtime,
  163     int waitfor)
  164 {
  165         struct m_ext2fs *fs;
  166         struct buf *bp;
  167         int error;
  168         struct timespec ts;
caddr_t cp;
  171         if (ITOV(ip)->v_mount->mnt_flag & MNT_RDONLY)
  172                 return (0);
getnanotime(&ts);
EXT2FS_ITIMES(ip,
atime ? atime : &ts,
mtime ? mtime : &ts);
  177         if ((ip->i_flag & IN_MODIFIED) == 0)
  178                 return (0);
ip->i_flag &= ~IN_MODIFIED;
fs = ip->i_e2fs;
error = bread(ip->i_devvp,
fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
  183                           (int)fs->e2fs_bsize, NOCRED, &bp);
  184         if (error) {
brelse(bp);
  186                 return (error);
  187         }
ip->i_flag &= ~(IN_MODIFIED);
cp = (caddr_t)bp->b_data +
  190             (ino_to_fsbo(fs, ip->i_number) * EXT2_DINODE_SIZE);
  192         /*
  193          * See note about 16-bit UID/GID limitation in ext2fs_vget(). Now
  194          * that we are about to write the inode, construct the split UID and
  195          * GID fields out of the two 32-bit fields we kept in memory.
  196          */
ip->i_e2fs_uid_low = (u_int16_t)ip->i_e2fs_uid;
ip->i_e2fs_gid_low = (u_int16_t)ip->i_e2fs_gid;
ip->i_e2fs_uid_high = ip->i_e2fs_uid >> 16;
ip->i_e2fs_gid_high = ip->i_e2fs_gid >> 16;
e2fs_isave(ip->i_e2din, (struct ext2fs_dinode *)cp);
  203         if (waitfor)
  204                 return (bwrite(bp));
  205         else {
bdwrite(bp);
  207                 return (0);
  208         }
  209 }
  211 #define SINGLE  0       /* index of single indirect block */
  212 #define DOUBLE  1       /* index of double indirect block */
  213 #define TRIPLE  2       /* index of triple indirect block */
  214 /*
  215  * Truncate the inode oip to at most length size, freeing the
  216  * disk blocks.
  217  */
  218 int
ext2fs_truncate(struct inode *oip, off_t length, int flags, struct ucred *cred)
  220 {
  221         struct vnode *ovp = ITOV(oip);
int32_t lastblock;
int32_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR];
int32_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
  225         struct m_ext2fs *fs;
  226         struct buf *bp;
  227         int offset, size, level;
  228         long count, nblocks, vflags, blocksreleased = 0;
  229         int i;
  230         int aflags, error, allerror;
off_t osize;
  233         if (length < 0)
  234                 return (EINVAL);
  236         if (ovp->v_type != VREG &&
ovp->v_type != VDIR &&
ovp->v_type != VLNK)
  239                 return (0);
  241         if (ovp->v_type == VLNK &&
  242                 (ext2fs_size(oip) < ovp->v_mount->mnt_maxsymlinklen ||
  243                  (ovp->v_mount->mnt_maxsymlinklen == 0 &&
oip->i_e2fs_nblock == 0))) {
  245 #ifdef DIAGNOSTIC
  246                 if (length != 0)
panic("ext2fs_truncate: partial truncate of symlink");
  248 #endif
bzero((char *)&oip->i_e2din->e2di_shortlink,
  250                         (u_int)ext2fs_size(oip));
  251                 (void)ext2fs_setsize(oip, 0);
oip->i_flag |= IN_CHANGE | IN_UPDATE;
  253                 return (ext2fs_update(oip, NULL, NULL, 1));
  254         }
  256         if (ext2fs_size(oip) == length) {
oip->i_flag |= IN_CHANGE | IN_UPDATE;
  258                 return (ext2fs_update(oip, NULL, NULL, 0));
  259         }
fs = oip->i_e2fs;
osize = ext2fs_size(oip);
  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 0 /* XXX */
  269                 if (length > fs->fs_maxfilesize)
  270                         return (EFBIG);
  271 #endif
offset = blkoff(fs, length - 1);
lbn = lblkno(fs, length - 1);
aflags = B_CLRBUF;
  275                 if (flags & IO_SYNC)
aflags |= B_SYNC;
error = ext2fs_buf_alloc(oip, lbn, offset + 1, cred, &bp,
aflags);
  279                 if (error)
  280                         return (error);
  281                 (void)ext2fs_setsize(oip, length);
uvm_vnp_setsize(ovp, length);
uvm_vnp_uncache(ovp);
  284                 if (aflags & B_SYNC)
bwrite(bp);
  286                 else
bawrite(bp);
oip->i_flag |= IN_CHANGE | IN_UPDATE;
  289                 return (ext2fs_update(oip, NULL, NULL, 1));
  290         }
  291         /*
  292          * Shorten the size of the file. If the file is not being
  293          * truncated to a block boundry, the contents of the
  294          * partial block following the end of the file must be
  295          * zero'ed in case it ever become accessible again because
  296          * of subsequent file growth.
  297          */
offset = blkoff(fs, length);
  299         if (offset == 0) {
  300                 (void)ext2fs_setsize(oip, length);
  301         } else {
lbn = lblkno(fs, length);
aflags = B_CLRBUF;
  304                 if (flags & IO_SYNC)
aflags |= B_SYNC;
error = ext2fs_buf_alloc(oip, lbn, offset, cred, &bp, 
aflags);
  308                 if (error)
  309                         return (error);
  310                 (void)ext2fs_setsize(oip, length);
size = fs->e2fs_bsize;
uvm_vnp_setsize(ovp, length);
uvm_vnp_uncache(ovp);
bzero(bp->b_data + offset, (u_int)(size - offset));
bp->b_bcount = size;
  316                 if (aflags & B_SYNC)
bwrite(bp);
  318                 else
bawrite(bp);
  320         }
  321         /*
  322          * Calculate index into inode's block list of
  323          * last direct and indirect blocks (if any)
  324          * which we want to keep.  Lastblock is -1 when
  325          * the file is truncated to 0.
  326          */
lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1;
lastiblock[SINGLE] = lastblock - NDADDR;
lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
nblocks = btodb(fs->e2fs_bsize);
  332         /*
  333          * Update file and block pointers on disk before we start freeing
  334          * blocks.  If we crash before free'ing blocks below, the blocks
  335          * will be returned to the free list.  lastiblock values are also
  336          * normalized to -1 for calls to ext2fs_indirtrunc below.
  337          */
memcpy((caddr_t)oldblks, (caddr_t)&oip->i_e2fs_blocks[0], sizeof oldblks);
  339         for (level = TRIPLE; level >= SINGLE; level--)
  340                 if (lastiblock[level] < 0) {
oip->i_e2fs_blocks[NDADDR + level] = 0;
lastiblock[level] = -1;
  343                 }
  344         for (i = NDADDR - 1; i > lastblock; i--)
oip->i_e2fs_blocks[i] = 0;
oip->i_flag |= IN_CHANGE | IN_UPDATE;
  347         if ((error = ext2fs_update(oip, NULL, NULL, 1)) != 0)
allerror = error;
  349         /*
  350          * Having written the new inode to disk, save its new configuration
  351          * and put back the old block pointers long enough to process them.
  352          * Note that we save the new block configuration so we can check it
  353          * when we are done.
  354          */
bcopy((caddr_t)&oip->i_e2fs_blocks[0], (caddr_t)newblks, sizeof newblks);
bcopy((caddr_t)oldblks, (caddr_t)&oip->i_e2fs_blocks[0], sizeof oldblks);
  357         (void)ext2fs_setsize(oip, osize);
vflags = ((length > 0) ? V_SAVE : 0) | V_SAVEMETA;
allerror = vinvalbuf(ovp, vflags, cred, curproc, 0, 0);
  361         /*
  362          * Indirect blocks first.
  363          */
indir_lbn[SINGLE] = -NDADDR;
indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) -1;
indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
  367         for (level = TRIPLE; level >= SINGLE; level--) {
bn = fs2h32(oip->i_e2fs_blocks[NDADDR + level]);
  369                 if (bn != 0) {
error = ext2fs_indirtrunc(oip, indir_lbn[level],
fsbtodb(fs, bn), lastiblock[level], level, &count);
  372                         if (error)
allerror = error;
blocksreleased += count;
  375                         if (lastiblock[level] < 0) {
oip->i_e2fs_blocks[NDADDR + level] = 0;
ext2fs_blkfree(oip, bn);
blocksreleased += nblocks;
  379                         }
  380                 }
  381                 if (lastiblock[level] >= 0)
  382                         goto done;
  383         }
  385         /*
  386          * All whole direct blocks or frags.
  387          */
  388         for (i = NDADDR - 1; i > lastblock; i--) {
bn = fs2h32(oip->i_e2fs_blocks[i]);
  390                 if (bn == 0)
  391                         continue;
oip->i_e2fs_blocks[i] = 0;
ext2fs_blkfree(oip, bn);
blocksreleased += btodb(fs->e2fs_bsize);
  395         }
done:
  398 #ifdef DIAGNOSTIC
  399         for (level = SINGLE; level <= TRIPLE; level++)
  400                 if (newblks[NDADDR + level] !=
oip->i_e2fs_blocks[NDADDR + level])
panic("ext2fs_truncate1");
  403         for (i = 0; i < NDADDR; i++)
  404                 if (newblks[i] != oip->i_e2fs_blocks[i])
panic("ext2fs_truncate2");
  406         if (length == 0 &&
  407             (!LIST_EMPTY(&ovp->v_cleanblkhd) ||
  408              !LIST_EMPTY(&ovp->v_dirtyblkhd)))
panic("ext2fs_truncate3");
  410 #endif /* DIAGNOSTIC */
  411         /*
  412          * Put back the real size.
  413          */
  414         (void)ext2fs_setsize(oip, length);
  415         if (blocksreleased >= oip->i_e2fs_nblock)
oip->i_e2fs_nblock = 0;
  417         else
oip->i_e2fs_nblock -= blocksreleased;
oip->i_flag |= IN_CHANGE;
  420         return (allerror);
  421 }
  423 /*
  424  * Release blocks associated with the inode ip and stored in the indirect
  425  * block bn.  Blocks are free'd in LIFO order up to (but not including)
  426  * lastbn.  If level is greater than SINGLE, the block is an indirect block
  427  * and recursive calls to indirtrunc must be used to cleanse other indirect
  428  * blocks.
  429  *
  430  * NB: triple indirect blocks are untested.
  431  */
  432 static int
ext2fs_indirtrunc(struct inode *ip, int32_t lbn, int32_t dbn, int32_t lastbn, int level, long *countp)
  434 {
  435         int i;
  436         struct buf *bp;
  437         struct m_ext2fs *fs = ip->i_e2fs;
int32_t *bap;
  439         struct vnode *vp;
int32_t *copy = NULL, nb, nlbn, last;
  441         long blkcount, factor;
  442         int nblocks, blocksreleased = 0;
  443         int error = 0, allerror = 0;
  445         /*
  446          * Calculate index in current block of last
  447          * block to be kept.  -1 indicates the entire
  448          * block so we need not calculate the index.
  449          */
factor = 1;
  451         for (i = SINGLE; i < level; i++)
factor *= NINDIR(fs);
last = lastbn;
  454         if (lastbn > 0)
last /= factor;
nblocks = btodb(fs->e2fs_bsize);
  457         /*
  458          * Get buffer of block pointers, zero those entries corresponding
  459          * to blocks to be free'd, and update on disk copy first.  Since
  460          * double(triple) indirect before single(double) indirect, calls
  461          * to bmap on these blocks will fail.  However, we already have
  462          * the on disk address, so we have to set the b_blkno field
  463          * explicitly instead of letting bread do everything for us.
  464          */
vp = ITOV(ip);
bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0);
  467         if (!(bp->b_flags & (B_DONE | B_DELWRI))) {
curproc->p_stats->p_ru.ru_inblock++;    /* pay for read */
bp->b_flags |= B_READ;
  470                 if (bp->b_bcount > bp->b_bufsize)
panic("ext2fs_indirtrunc: bad buffer size");
bp->b_blkno = dbn;
VOP_STRATEGY(bp);
error = biowait(bp);
  475         }
  476         if (error) {
brelse(bp);
  478                 *countp = 0;
  479                 return (error);
  480         }
bap = (int32_t *)bp->b_data;
  483         if (lastbn >= 0) {
MALLOC(copy, int32_t *, fs->e2fs_bsize, M_TEMP, M_WAITOK);
memcpy((caddr_t)copy, (caddr_t)bap, (u_int)fs->e2fs_bsize);
memset((caddr_t)&bap[last + 1], 0,
  487                         (u_int)(NINDIR(fs) - (last + 1)) * sizeof (u_int32_t));
error = bwrite(bp);
  489                 if (error)
allerror = error;
bap = copy;
  492         }
  494         /*
  495          * Recursively free totally unused blocks.
  496          */
  497         for (i = NINDIR(fs) - 1,
nlbn = lbn + 1 - i * factor; i > last;
i--, nlbn += factor) {
nb = fs2h32(bap[i]);
  501                 if (nb == 0)
  502                         continue;
  503                 if (level > SINGLE) {
error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
  505                                                    (int32_t)-1, level - 1,
  506                                                    &blkcount);
  507                         if (error)
allerror = error;
blocksreleased += blkcount;
  510                 }
ext2fs_blkfree(ip, nb);
blocksreleased += nblocks;
  513         }
  515         /*
  516          * Recursively free last partial block.
  517          */
  518         if (level > SINGLE && lastbn >= 0) {
last = lastbn % factor;
nb = fs2h32(bap[i]);
  521                 if (nb != 0) {
error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
last, level - 1, &blkcount);
  524                         if (error)
allerror = error;
blocksreleased += blkcount;
  527                 }
  528         }
  530         if (copy != NULL) {
FREE(copy, M_TEMP);
  532         } else {
bp->b_flags |= B_INVAL;
brelse(bp);
  535         }
  537         *countp = blocksreleased;
  538         return (allerror);
  539 }