int
fsinit1(time_t utime, mode_t mfsmode, uid_t mfsuid, gid_t mfsgid)
{
union dinode node;
/*
* Initialize the node
*/
memset(&node, 0, sizeof(node));
node.dp1.di_atime = utime;
node.dp1.di_mtime = utime;
node.dp1.di_ctime = utime;
/*
* Create the root directory.
*/
if (mfs) {
node.dp1.di_mode = IFDIR | mfsmode;
node.dp1.di_uid = mfsuid;
node.dp1.di_gid = mfsgid;
} else {
node.dp1.di_mode = IFDIR | UMASK;
node.dp1.di_uid = geteuid();
node.dp1.di_gid = getegid();
}
node.dp1.di_nlink = PREDEFDIR;
if (Oflag == 0)
node.dp1.di_size = makedir((struct direct *)oroot_dir,
PREDEFDIR);
else
node.dp1.di_size = makedir(root_dir, PREDEFDIR);
node.dp1.di_db[0] = alloc(sblock.fs_fsize, node.dp1.di_mode);
if (node.dp1.di_db[0] == 0)
return (1);
node.dp1.di_blocks = btodb(fragroundup(&sblock, node.dp1.di_size));
wtfs(fsbtodb(&sblock, node.dp1.di_db[0]), sblock.fs_fsize, iobuf);
iput(&node, ROOTINO);
#ifdef notyet
/*
* Create the .snap directory.
*/
node.dp1.di_mode |= 020;
node.dp1.di_gid = gid;
node.dp1.di_nlink = SNAPLINKCNT;
node.dp1.di_size = makedir(snap_dir, SNAPLINKCNT);
node.dp1.di_db[0] = alloc(sblock.fs_fsize, node.dp1.di_mode);
if (node.dp1.di_db[0] == 0)
return (1);
node.dp1.di_blocks = btodb(fragroundup(&sblock, node.dp1.di_size));
wtfs(fsbtodb(&sblock, node.dp1.di_db[0]), sblock.fs_fsize, iobuf);
iput(&node, ROOTINO + 1);
#endif
return (0);
}
static int
calc_num_blocks(struct inode *inode)
{
int l1_indir, l2_indir, l3_indir;
struct fs *fs = inode->i_fs;
int nfrags, numblocks = howmany(inode->i_size, fs->fs_bsize);
if (numblocks < NDADDR) {
nfrags = numfrags(fs, fragroundup(fs, inode->i_size));
} else {
int nindirs = fs->fs_nindir;
/* Calculate how many indirects do we need to hold these many blocks */
l1_indir = howmany(numblocks - NDADDR, nindirs);
numblocks += l1_indir;
l2_indir = howmany(l1_indir - 1, nindirs);
numblocks += l2_indir;
if (l2_indir) {
l3_indir = howmany(l2_indir - 1, nindirs);
numblocks += l3_indir;
}
nfrags = numblocks * fs->fs_frag;
}
return nfrags * (fs->fs_fsize >> DEV_BSHIFT);
}
static int
virtualblocks(struct fs *super, union dinode *dp)
{
off_t nblk, sz;
sz = DIP(super, dp, di_size);
#ifdef COMPAT
if (lblkno(super, sz) >= NDADDR) {
nblk = blkroundup(super, sz);
if (sz == nblk)
nblk += super->fs_bsize;
}
return sz / 1024;
#else /* COMPAT */
if (lblkno(super, sz) >= NDADDR) {
nblk = blkroundup(super, sz);
sz = lblkno(super, nblk);
sz = howmany(sz - NDADDR, NINDIR(super));
while (sz > 0) {
nblk += sz * super->fs_bsize;
/* One block on this level is in the inode itself */
sz = howmany(sz - 1, NINDIR(super));
}
} else
nblk = fragroundup(super, sz);
return nblk / DEV_BSIZE;
#endif /* COMPAT */
}
static int virtualblocks(struct fs *super, union dinode *dp)
{
off_t nblk, sz;
sz = DIP(super, dp, di_size);
#ifdef COMPAT
if (lblkno(super,sz) >= NDADDR) {
nblk = blkroundup(super,sz);
if (sz == nblk)
nblk += super->fs_bsize;
}
return sz / 1024;
#else /* COMPAT */
if (lblkno(super,sz) >= NDADDR) {
nblk = blkroundup(super,sz);
sz = lblkno(super,nblk);
sz = (sz - NDADDR + NINDIR(super) - 1) / NINDIR(super);
while (sz > 0) {
nblk += sz * super->fs_bsize;
/* sz - 1 rounded up */
sz = (sz - 1 + NINDIR(super) - 1) / NINDIR(super);
}
} else
nblk = fragroundup(super,sz);
return nblk / 512;
#endif /* COMPAT */
}
static int
marshal(const char *name)
{
struct fs *fs;
fs = &disk.d_fs;
printf("# newfs command for %s (%s)\n", name, disk.d_name);
printf("newfs ");
if (fs->fs_volname[0] != '\0')
printf("-L %s ", fs->fs_volname);
printf("-O %d ", disk.d_ufs);
if (fs->fs_flags & FS_DOSOFTDEP)
printf("-U ");
printf("-a %d ", fs->fs_maxcontig);
printf("-b %d ", fs->fs_bsize);
/* -c is dumb */
printf("-d %d ", fs->fs_maxbsize);
printf("-e %d ", fs->fs_maxbpg);
printf("-f %d ", fs->fs_fsize);
printf("-g %d ", fs->fs_avgfilesize);
printf("-h %d ", fs->fs_avgfpdir);
printf("-i %jd ", fragroundup(fs, lblktosize(fs, fragstoblks(fs,
fs->fs_fpg)) / fs->fs_ipg));
if (fs->fs_flags & FS_SUJ)
printf("-j ");
if (fs->fs_flags & FS_GJOURNAL)
printf("-J ");
printf("-k %jd ", fs->fs_metaspace);
if (fs->fs_flags & FS_MULTILABEL)
printf("-l ");
printf("-m %d ", fs->fs_minfree);
/* -n unimplemented */
printf("-o ");
switch (fs->fs_optim) {
case FS_OPTSPACE:
printf("space ");
break;
case FS_OPTTIME:
printf("time ");
break;
default:
printf("unknown ");
break;
}
/* -p..r unimplemented */
printf("-s %jd ", (intmax_t)fsbtodb(fs, fs->fs_size));
if (fs->fs_flags & FS_TRIM)
printf("-t ");
printf("%s ", disk.d_name);
printf("\n");
return 0;
}
static int ufs_new_dir_block(uufsd_t *ufs, ino_t dir_ino,
struct ufs_vnode *parent, char **block)
{
struct direct *dir = NULL;
int retval;
char *buf;
int rec_len;
struct fs *fs = &ufs->d_fs;
int dirsize = DIRBLKSIZ;
int blocksize = fragroundup(fs, dirsize);
retval = ufs_get_mem(blocksize, &buf);
if (retval)
return retval;
memset(buf, 0, blocksize);
dir = (struct direct *) buf;
//retval = ufs_set_rec_len(ufs, dirsize, dir);
//if (retval)
//return retval;
if (dir_ino) {
/*
* Set up entry for '.'
*/
dir->d_ino = dir_ino;
dir->d_namlen = 1;
dir->d_name[0] = '.';
dir->d_type = DT_DIR;
rec_len = dirsize - UFS_DIR_REC_LEN(1);
dir->d_reclen = UFS_DIR_REC_LEN(1);
/*
* Set up entry for '..'
*/
dir = (struct direct *) (buf + dir->d_reclen);
//retval = ufs_set_rec_len(ufs, rec_len, dir);
//if (retval)
//return retval;
dir->d_ino = vnode2inode(parent)->i_number;
dir->d_namlen = 2;
dir->d_name[0] = '.';
dir->d_name[1] = '.';
dir->d_type = DT_DIR;
dir->d_reclen = rec_len;
}
*block = buf;
return 0;
}
void
ffs_gop_size(struct vnode *vp, off_t size, off_t *eobp, int flags)
{
struct inode *ip = VTOI(vp);
struct fs *fs = ip->i_fs;
daddr_t olbn, nlbn;
olbn = lblkno(fs, ip->i_size);
nlbn = lblkno(fs, size);
if (nlbn < NDADDR && olbn <= nlbn) {
*eobp = fragroundup(fs, size);
} else {
*eobp = blkroundup(fs, size);
}
}
/*
* Print the block pointers for one inode.
*/
static void
printblocks(ino_t inum, union dinode *dp)
{
char *bufp;
int i, nfrags;
long ndb, offset;
ufs2_daddr_t blkno;
printf("Blocks for inode %d:\n", inum);
printf("Direct blocks:\n");
ndb = howmany(DIP(dp, di_size), sblock.fs_bsize);
for (i = 0; i < NDADDR && i < ndb; i++) {
if (i > 0)
printf(", ");
blkno = DIP(dp, di_db[i]);
printf("%jd", (intmax_t)blkno);
}
if (ndb <= NDADDR) {
offset = blkoff(&sblock, DIP(dp, di_size));
if (offset != 0) {
nfrags = numfrags(&sblock, fragroundup(&sblock, offset));
printf(" (%d frag%s)", nfrags, nfrags > 1? "s": "");
}
}
putchar('\n');
if (ndb <= NDADDR)
return;
bufp = malloc((unsigned int)sblock.fs_bsize);
if (bufp == 0)
errx(EEXIT, "cannot allocate indirect block buffer");
printf("Indirect blocks:\n");
for (i = 0; i < NIADDR; i++)
printindir(DIP(dp, di_ib[i]), i, bufp);
free(bufp);
}
/*
* Print the block pointers for one inode.
*/
static void
printblocks(ino_t inum, struct ufs1_dinode *dp)
{
char *bufp;
int i, nfrags;
long ndb, offset;
printf("Blocks for inode %d:\n", inum);
printf("Direct blocks:\n");
ndb = howmany(dp->di_size, sblock.fs_bsize);
for (i = 0; i < NDADDR; i++) {
if (dp->di_db[i] == 0) {
putchar('\n');
return;
}
if (i > 0)
printf(", ");
printf("%d", dp->di_db[i]);
if (--ndb == 0 && (offset = blkoff(&sblock, dp->di_size)) != 0) {
nfrags = numfrags(&sblock, fragroundup(&sblock, offset));
printf(" (%d frag%s)", nfrags, nfrags > 1? "s": "");
}
}
putchar('\n');
if (dp->di_ib[0] == 0)
return;
bufp = malloc((unsigned int)sblock.fs_bsize);
if (bufp == NULL)
errx(EEXIT, "cannot allocate indirect block buffer");
printf("Indirect blocks:\n");
for (i = 0; i < NIADDR; i++)
if (printindir(dp->di_ib[i], i, bufp) == 0)
break;
free(bufp);
}
/*
* Truncate the inode oip to at most length size, freeing the
* disk blocks.
*/
int
ffs_truncate(struct vnode *vp, off_t length, int flags, struct ucred *cred)
{
struct vnode *ovp = vp;
ufs_daddr_t lastblock;
struct inode *oip;
ufs_daddr_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR];
ufs_daddr_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
struct fs *fs;
struct buf *bp;
int offset, size, level;
long count, nblocks, blocksreleased = 0;
int i;
int aflags, error, allerror;
off_t osize;
oip = VTOI(ovp);
fs = oip->i_fs;
if (length < 0)
return (EINVAL);
if (length > fs->fs_maxfilesize)
return (EFBIG);
if (ovp->v_type == VLNK &&
(oip->i_size < ovp->v_mount->mnt_maxsymlinklen || oip->i_din.di_blocks == 0)) {
#ifdef DIAGNOSTIC
if (length != 0)
panic("ffs_truncate: partial truncate of symlink");
#endif /* DIAGNOSTIC */
bzero((char *)&oip->i_shortlink, (uint)oip->i_size);
oip->i_size = 0;
oip->i_flag |= IN_CHANGE | IN_UPDATE;
return (ffs_update(ovp, 1));
}
if (oip->i_size == length) {
oip->i_flag |= IN_CHANGE | IN_UPDATE;
return (ffs_update(ovp, 0));
}
if (fs->fs_ronly)
panic("ffs_truncate: read-only filesystem");
#ifdef QUOTA
error = ufs_getinoquota(oip);
if (error)
return (error);
#endif
ovp->v_lasta = ovp->v_clen = ovp->v_cstart = ovp->v_lastw = 0;
if (DOINGSOFTDEP(ovp)) {
if (length > 0 || softdep_slowdown(ovp)) {
/*
* If a file is only partially truncated, then
* we have to clean up the data structures
* describing the allocation past the truncation
* point. Finding and deallocating those structures
* is a lot of work. Since partial truncation occurs
* rarely, we solve the problem by syncing the file
* so that it will have no data structures left.
*/
if ((error = VOP_FSYNC(ovp, MNT_WAIT, 0)) != 0)
return (error);
} else {
#ifdef QUOTA
(void) ufs_chkdq(oip, -oip->i_blocks, NOCRED, 0);
#endif
softdep_setup_freeblocks(oip, length);
vinvalbuf(ovp, 0, 0, 0);
nvnode_pager_setsize(ovp, 0, fs->fs_bsize, 0);
oip->i_flag |= IN_CHANGE | IN_UPDATE;
return (ffs_update(ovp, 0));
}
}
osize = oip->i_size;
/*
* Lengthen the size of the file. We must ensure that the
* last byte of the file is allocated. Since the smallest
* value of osize is 0, length will be at least 1.
*
* nvextendbuf() only breads the old buffer. The blocksize
* of the new buffer must be specified so it knows how large
* to make the VM object.
*/
if (osize < length) {
nvextendbuf(vp, osize, length,
blkoffsize(fs, oip, osize), /* oblksize */
blkoffresize(fs, length), /* nblksize */
blkoff(fs, osize),
blkoff(fs, length),
0);
aflags = B_CLRBUF;
if (flags & IO_SYNC)
aflags |= B_SYNC;
/* BALLOC will reallocate the fragment at the old EOF */
error = VOP_BALLOC(ovp, length - 1, 1, cred, aflags, &bp);
if (error)
return (error);
oip->i_size = length;
if (bp->b_bufsize == fs->fs_bsize)
bp->b_flags |= B_CLUSTEROK;
if (aflags & B_SYNC)
bwrite(bp);
else
bawrite(bp);
oip->i_flag |= IN_CHANGE | IN_UPDATE;
return (ffs_update(ovp, 1));
}
/*
* Shorten the size of the file.
*
* NOTE: The block size specified in nvtruncbuf() is the blocksize
* of the buffer containing length prior to any reallocation
* of the block.
*/
allerror = nvtruncbuf(ovp, length, blkoffsize(fs, oip, length),
blkoff(fs, length), 0);
offset = blkoff(fs, length);
if (offset == 0) {
oip->i_size = length;
} else {
lbn = lblkno(fs, length);
aflags = B_CLRBUF;
if (flags & IO_SYNC)
aflags |= B_SYNC;
error = VOP_BALLOC(ovp, length - 1, 1, cred, aflags, &bp);
if (error)
return (error);
/*
* When we are doing soft updates and the UFS_BALLOC
* above fills in a direct block hole with a full sized
* block that will be truncated down to a fragment below,
* we must flush out the block dependency with an FSYNC
* so that we do not get a soft updates inconsistency
* when we create the fragment below.
*
* nvtruncbuf() may have re-dirtied the underlying block
* as part of its truncation zeroing code. To avoid a
* 'locking against myself' panic in the second fsync we
* can simply undirty the bp since the redirtying was
* related to areas of the buffer that we are going to
* throw away anyway, and we will b*write() the remainder
* anyway down below.
*/
if (DOINGSOFTDEP(ovp) && lbn < NDADDR &&
fragroundup(fs, blkoff(fs, length)) < fs->fs_bsize) {
bundirty(bp);
error = VOP_FSYNC(ovp, MNT_WAIT, 0);
if (error) {
bdwrite(bp);
return (error);
}
}
oip->i_size = length;
size = blksize(fs, oip, lbn);
#if 0
/* remove - nvtruncbuf deals with this */
if (ovp->v_type != VDIR)
bzero((char *)bp->b_data + offset,
(uint)(size - offset));
#endif
/* Kirk's code has reallocbuf(bp, size, 1) here */
allocbuf(bp, size);
if (bp->b_bufsize == fs->fs_bsize)
bp->b_flags |= B_CLUSTEROK;
if (aflags & B_SYNC)
bwrite(bp);
else
bawrite(bp);
}
/*
* Calculate index into inode's block list of
* last direct and indirect blocks (if any)
* which we want to keep. Lastblock is -1 when
* the file is truncated to 0.
*/
lastblock = lblkno(fs, length + fs->fs_bsize - 1) - 1;
lastiblock[SINGLE] = lastblock - NDADDR;
lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
nblocks = btodb(fs->fs_bsize);
/*
* Update file and block pointers on disk before we start freeing
* blocks. If we crash before free'ing blocks below, the blocks
* will be returned to the free list. lastiblock values are also
* normalized to -1 for calls to ffs_indirtrunc below.
*/
bcopy((caddr_t)&oip->i_db[0], (caddr_t)oldblks, sizeof oldblks);
for (level = TRIPLE; level >= SINGLE; level--)
if (lastiblock[level] < 0) {
oip->i_ib[level] = 0;
lastiblock[level] = -1;
}
for (i = NDADDR - 1; i > lastblock; i--)
oip->i_db[i] = 0;
oip->i_flag |= IN_CHANGE | IN_UPDATE;
error = ffs_update(ovp, 1);
if (error && allerror == 0)
allerror = error;
/*
* Having written the new inode to disk, save its new configuration
* and put back the old block pointers long enough to process them.
* Note that we save the new block configuration so we can check it
* when we are done.
*/
bcopy((caddr_t)&oip->i_db[0], (caddr_t)newblks, sizeof newblks);
bcopy((caddr_t)oldblks, (caddr_t)&oip->i_db[0], sizeof oldblks);
oip->i_size = osize;
if (error && allerror == 0)
allerror = error;
/*
* Indirect blocks first.
*/
indir_lbn[SINGLE] = -NDADDR;
indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1;
indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
for (level = TRIPLE; level >= SINGLE; level--) {
bn = oip->i_ib[level];
if (bn != 0) {
error = ffs_indirtrunc(oip, indir_lbn[level],
fsbtodb(fs, bn), lastiblock[level], level, &count);
if (error)
allerror = error;
blocksreleased += count;
if (lastiblock[level] < 0) {
oip->i_ib[level] = 0;
ffs_blkfree(oip, bn, fs->fs_bsize);
blocksreleased += nblocks;
}
}
if (lastiblock[level] >= 0)
goto done;
}
/*
* All whole direct blocks or frags.
*/
for (i = NDADDR - 1; i > lastblock; i--) {
long bsize;
bn = oip->i_db[i];
if (bn == 0)
continue;
oip->i_db[i] = 0;
bsize = blksize(fs, oip, i);
ffs_blkfree(oip, bn, bsize);
blocksreleased += btodb(bsize);
}
if (lastblock < 0)
goto done;
/*
* Finally, look for a change in size of the
* last direct block; release any frags.
*/
bn = oip->i_db[lastblock];
if (bn != 0) {
long oldspace, newspace;
/*
* Calculate amount of space we're giving
* back as old block size minus new block size.
*/
oldspace = blksize(fs, oip, lastblock);
oip->i_size = length;
newspace = blksize(fs, oip, lastblock);
if (newspace == 0)
panic("ffs_truncate: newspace");
if (oldspace - newspace > 0) {
/*
* Block number of space to be free'd is
* the old block # plus the number of frags
* required for the storage we're keeping.
*/
bn += numfrags(fs, newspace);
ffs_blkfree(oip, bn, oldspace - newspace);
blocksreleased += btodb(oldspace - newspace);
}
}
done:
#ifdef DIAGNOSTIC
for (level = SINGLE; level <= TRIPLE; level++)
if (newblks[NDADDR + level] != oip->i_ib[level])
panic("ffs_truncate1");
for (i = 0; i < NDADDR; i++)
if (newblks[i] != oip->i_db[i])
panic("ffs_truncate2");
if (length == 0 && !RB_EMPTY(&ovp->v_rbdirty_tree))
panic("ffs_truncate3");
#endif /* DIAGNOSTIC */
/*
* Put back the real size.
*/
oip->i_size = length;
oip->i_blocks -= blocksreleased;
if (oip->i_blocks < 0) /* sanity */
oip->i_blocks = 0;
oip->i_flag |= IN_CHANGE;
#ifdef QUOTA
(void) ufs_chkdq(oip, -blocksreleased, NOCRED, 0);
#endif
return (allerror);
}
void
fsinit(time_t utime)
{
union dinode node;
struct group *grp;
gid_t gid;
int entries;
memset(&node, 0, sizeof node);
if ((grp = getgrnam("operator")) != NULL) {
gid = grp->gr_gid;
} else {
warnx("Cannot retrieve operator gid, using gid 0.");
gid = 0;
}
entries = (nflag) ? ROOTLINKCNT - 1: ROOTLINKCNT;
if (sblock.fs_magic == FS_UFS1_MAGIC) {
/*
* initialize the node
*/
node.dp1.di_atime = utime;
node.dp1.di_mtime = utime;
node.dp1.di_ctime = utime;
/*
* create the root directory
*/
node.dp1.di_mode = IFDIR | UMASK;
node.dp1.di_nlink = entries;
node.dp1.di_size = makedir(root_dir, entries);
node.dp1.di_db[0] = alloc(sblock.fs_fsize, node.dp1.di_mode);
node.dp1.di_blocks =
btodb(fragroundup(&sblock, node.dp1.di_size));
wtfs(fsbtodb(&sblock, node.dp1.di_db[0]), sblock.fs_fsize,
iobuf);
iput(&node, ROOTINO);
if (!nflag) {
/*
* create the .snap directory
*/
node.dp1.di_mode |= 020;
node.dp1.di_gid = gid;
node.dp1.di_nlink = SNAPLINKCNT;
node.dp1.di_size = makedir(snap_dir, SNAPLINKCNT);
node.dp1.di_db[0] =
alloc(sblock.fs_fsize, node.dp1.di_mode);
node.dp1.di_blocks =
btodb(fragroundup(&sblock, node.dp1.di_size));
wtfs(fsbtodb(&sblock, node.dp1.di_db[0]),
sblock.fs_fsize, iobuf);
iput(&node, ROOTINO + 1);
}
} else {
/*
* initialize the node
*/
node.dp2.di_atime = utime;
node.dp2.di_mtime = utime;
node.dp2.di_ctime = utime;
node.dp2.di_birthtime = utime;
/*
* create the root directory
*/
node.dp2.di_mode = IFDIR | UMASK;
node.dp2.di_nlink = entries;
node.dp2.di_size = makedir(root_dir, entries);
node.dp2.di_db[0] = alloc(sblock.fs_fsize, node.dp2.di_mode);
node.dp2.di_blocks =
btodb(fragroundup(&sblock, node.dp2.di_size));
wtfs(fsbtodb(&sblock, node.dp2.di_db[0]), sblock.fs_fsize,
iobuf);
iput(&node, ROOTINO);
if (!nflag) {
/*
* create the .snap directory
*/
node.dp2.di_mode |= 020;
node.dp2.di_gid = gid;
node.dp2.di_nlink = SNAPLINKCNT;
node.dp2.di_size = makedir(snap_dir, SNAPLINKCNT);
node.dp2.di_db[0] =
alloc(sblock.fs_fsize, node.dp2.di_mode);
node.dp2.di_blocks =
btodb(fragroundup(&sblock, node.dp2.di_size));
wtfs(fsbtodb(&sblock, node.dp2.di_db[0]),
sblock.fs_fsize, iobuf);
iput(&node, ROOTINO + 1);
}
}
}
static int
ffs_balloc_ufs1(struct inode *ip, off_t offset, int bufsize, struct buf **bpp)
{
makefs_daddr_t lbn, lastlbn;
int size;
int32_t nb;
struct buf *bp, *nbp;
struct fs *fs = ip->i_fs;
struct indir indirs[UFS_NIADDR + 2];
makefs_daddr_t newb, pref;
int32_t *bap;
int osize, nsize, num, i, error;
int32_t *allocblk, allociblk[UFS_NIADDR + 1];
int32_t *allocib;
const int needswap = UFS_FSNEEDSWAP(fs);
lbn = lblkno(fs, offset);
size = blkoff(fs, offset) + bufsize;
if (bpp != NULL) {
*bpp = NULL;
}
assert(size <= fs->fs_bsize);
if (lbn < 0)
return (EFBIG);
/*
* If the next write will extend the file into a new block,
* and the file is currently composed of a fragment
* this fragment has to be extended to be a full block.
*/
lastlbn = lblkno(fs, ip->i_ffs1_size);
if (lastlbn < UFS_NDADDR && lastlbn < lbn) {
nb = lastlbn;
osize = blksize(fs, ip, nb);
if (osize < fs->fs_bsize && osize > 0) {
warnx("need to ffs_realloccg; not supported!");
abort();
}
}
/*
* The first UFS_NDADDR blocks are direct blocks
*/
if (lbn < UFS_NDADDR) {
nb = ufs_rw32(ip->i_ffs1_db[lbn], needswap);
if (nb != 0 && ip->i_ffs1_size >=
(uint64_t)lblktosize(fs, lbn + 1)) {
/*
* The block is an already-allocated direct block
* and the file already extends past this block,
* thus this must be a whole block.
* Just read the block (if requested).
*/
if (bpp != NULL) {
error = bread(ip->i_devvp, lbn, fs->fs_bsize,
NULL, bpp);
if (error) {
brelse(*bpp);
return (error);
}
}
return (0);
}
if (nb != 0) {
/*
* Consider need to reallocate a fragment.
*/
osize = fragroundup(fs, blkoff(fs, ip->i_ffs1_size));
nsize = fragroundup(fs, size);
if (nsize <= osize) {
/*
* The existing block is already
* at least as big as we want.
* Just read the block (if requested).
*/
if (bpp != NULL) {
error = bread(ip->i_devvp, lbn, osize,
NULL, bpp);
if (error) {
brelse(*bpp);
return (error);
}
}
return 0;
} else {
warnx("need to ffs_realloccg; not supported!");
abort();
}
} else {
/*
* the block was not previously allocated,
* allocate a new block or fragment.
*/
if (ip->i_ffs1_size < (uint64_t)lblktosize(fs, lbn + 1))
nsize = fragroundup(fs, size);
else
nsize = fs->fs_bsize;
error = ffs_alloc(ip, lbn,
ffs_blkpref_ufs1(ip, lbn, (int)lbn,
&ip->i_ffs1_db[0]),
nsize, &newb);
if (error)
return (error);
if (bpp != NULL) {
bp = getblk(ip->i_devvp, lbn, nsize, 0, 0, 0);
bp->b_blkno = fsbtodb(fs, newb);
clrbuf(bp);
*bpp = bp;
}
}
ip->i_ffs1_db[lbn] = ufs_rw32((int32_t)newb, needswap);
return (0);
}
/*
* Determine the number of levels of indirection.
*/
pref = 0;
if ((error = ufs_getlbns(ip, lbn, indirs, &num)) != 0)
return (error);
if (num < 1) {
warnx("ffs_balloc: ufs_getlbns returned indirect block");
abort();
}
/*
* Fetch the first indirect block allocating if necessary.
*/
--num;
nb = ufs_rw32(ip->i_ffs1_ib[indirs[0].in_off], needswap);
allocib = NULL;
allocblk = allociblk;
if (nb == 0) {
pref = ffs_blkpref_ufs1(ip, lbn, 0, (int32_t *)0);
error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, &newb);
if (error)
return error;
nb = newb;
*allocblk++ = nb;
bp = getblk(ip->i_devvp, indirs[1].in_lbn, fs->fs_bsize, 0, 0, 0);
bp->b_blkno = fsbtodb(fs, nb);
clrbuf(bp);
/*
* Write synchronously so that indirect blocks
* never point at garbage.
*/
if ((error = bwrite(bp)) != 0)
return error;
allocib = &ip->i_ffs1_ib[indirs[0].in_off];
*allocib = ufs_rw32((int32_t)nb, needswap);
}
/*
* Fetch through the indirect blocks, allocating as necessary.
*/
for (i = 1;;) {
error = bread(ip->i_devvp, indirs[i].in_lbn, fs->fs_bsize,
NULL, &bp);
if (error) {
brelse(bp);
return error;
}
bap = (int32_t *)bp->b_data;
nb = ufs_rw32(bap[indirs[i].in_off], needswap);
if (i == num)
break;
i++;
if (nb != 0) {
brelse(bp);
continue;
}
if (pref == 0)
pref = ffs_blkpref_ufs1(ip, lbn, 0, (int32_t *)0);
error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, &newb);
if (error) {
brelse(bp);
return error;
}
nb = newb;
*allocblk++ = nb;
nbp = getblk(ip->i_devvp, indirs[i].in_lbn, fs->fs_bsize, 0, 0, 0);
nbp->b_blkno = fsbtodb(fs, nb);
clrbuf(nbp);
/*
* Write synchronously so that indirect blocks
* never point at garbage.
*/
if ((error = bwrite(nbp)) != 0) {
brelse(bp);
return error;
}
bap[indirs[i - 1].in_off] = ufs_rw32(nb, needswap);
bwrite(bp);
}
/*
* Get the data block, allocating if necessary.
*/
if (nb == 0) {
pref = ffs_blkpref_ufs1(ip, lbn, indirs[num].in_off, &bap[0]);
error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, &newb);
if (error) {
brelse(bp);
return error;
}
nb = newb;
*allocblk++ = nb;
if (bpp != NULL) {
nbp = getblk(ip->i_devvp, lbn, fs->fs_bsize, 0, 0, 0);
nbp->b_blkno = fsbtodb(fs, nb);
clrbuf(nbp);
*bpp = nbp;
}
bap[indirs[num].in_off] = ufs_rw32(nb, needswap);
/*
* If required, write synchronously, otherwise use
* delayed write.
*/
bwrite(bp);
return (0);
}
brelse(bp);
if (bpp != NULL) {
error = bread(ip->i_devvp, lbn, (int)fs->fs_bsize, NULL, &nbp);
if (error) {
brelse(nbp);
return error;
}
*bpp = nbp;
}
return (0);
}
/*
* ffs_blkalloc allocates a disk block for ffs_pageout(), as a consequence
* it does no buf_breads (that could lead to deadblock as the page may be already
* marked busy as it is being paged out. Also important to note that we are not
* growing the file in pageouts. So ip->i_size cannot increase by this call
* due to the way UBC works.
* This code is derived from ffs_balloc and many cases of that are dealt
* in ffs_balloc are not applicable here
* Do not call with B_CLRBUF flags as this should only be called only
* from pageouts
*/
ffs_blkalloc(
struct inode *ip,
ufs_daddr_t lbn,
int size,
kauth_cred_t cred,
int flags)
{
register struct fs *fs;
register ufs_daddr_t nb;
struct buf *bp, *nbp;
struct vnode *vp = ITOV(ip);
struct indir indirs[NIADDR + 2];
ufs_daddr_t newb, *bap, pref;
int deallocated, osize, nsize, num, i, error;
ufs_daddr_t *allocib, *blkp, *allocblk, allociblk[NIADDR + 1];
int devBlockSize=0;
struct mount *mp=vp->v_mount;
#if REV_ENDIAN_FS
int rev_endian=(mp->mnt_flag & MNT_REVEND);
#endif /* REV_ENDIAN_FS */
fs = ip->i_fs;
if(size > fs->fs_bsize)
panic("ffs_blkalloc: too large for allocation");
/*
* If the next write will extend the file into a new block,
* and the file is currently composed of a fragment
* this fragment has to be extended to be a full block.
*/
nb = lblkno(fs, ip->i_size);
if (nb < NDADDR && nb < lbn) {
panic("ffs_blkalloc():cannot extend file: i_size %d, lbn %d", ip->i_size, lbn);
}
/*
* The first NDADDR blocks are direct blocks
*/
if (lbn < NDADDR) {
nb = ip->i_db[lbn];
if (nb != 0 && ip->i_size >= (lbn + 1) * fs->fs_bsize) {
/* TBD: trivial case; the block is already allocated */
return (0);
}
if (nb != 0) {
/*
* Consider need to reallocate a fragment.
*/
osize = fragroundup(fs, blkoff(fs, ip->i_size));
nsize = fragroundup(fs, size);
if (nsize > osize) {
panic("ffs_allocblk: trying to extend a fragment");
}
return(0);
} else {
if (ip->i_size < (lbn + 1) * fs->fs_bsize)
nsize = fragroundup(fs, size);
else
nsize = fs->fs_bsize;
error = ffs_alloc(ip, lbn,
ffs_blkpref(ip, lbn, (int)lbn, &ip->i_db[0]),
nsize, cred, &newb);
if (error)
return (error);
ip->i_db[lbn] = newb;
ip->i_flag |= IN_CHANGE | IN_UPDATE;
return (0);
}
}
/*
* Determine the number of levels of indirection.
*/
pref = 0;
if (error = ufs_getlbns(vp, lbn, indirs, &num))
return(error);
if(num == 0) {
panic("ffs_blkalloc: file with direct blocks only");
}
/*
* Fetch the first indirect block allocating if necessary.
*/
--num;
nb = ip->i_ib[indirs[0].in_off];
allocib = NULL;
allocblk = allociblk;
if (nb == 0) {
pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0);
if (error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize,
cred, &newb))
return (error);
nb = newb;
*allocblk++ = nb;
bp = buf_getblk(vp, (daddr64_t)((unsigned)(indirs[1].in_lbn)), fs->fs_bsize, 0, 0, BLK_META);
buf_setblkno(bp, (daddr64_t)((unsigned)fsbtodb(fs, nb)));
buf_clear(bp);
/*
* Write synchronously conditional on mount flags.
*/
if ((vp)->v_mount->mnt_flag & MNT_ASYNC) {
error = 0;
buf_bdwrite(bp);
} else if (error = buf_bwrite(bp)) {
goto fail;
}
allocib = &ip->i_ib[indirs[0].in_off];
*allocib = nb;
ip->i_flag |= IN_CHANGE | IN_UPDATE;
}
/*
* Fetch through the indirect blocks, allocating as necessary.
*/
for (i = 1;;) {
error = (int)buf_meta_bread(vp, (daddr64_t)((unsigned)(indirs[i].in_lbn)), (int)fs->fs_bsize, NOCRED, &bp);
if (error) {
buf_brelse(bp);
goto fail;
}
bap = (ufs_daddr_t *)buf_dataptr(bp);
#if REV_ENDIAN_FS
if (rev_endian)
nb = OSSwapInt32(bap[indirs[i].in_off]);
else {
#endif /* REV_ENDIAN_FS */
nb = bap[indirs[i].in_off];
#if REV_ENDIAN_FS
}
#endif /* REV_ENDIAN_FS */
if (i == num)
break;
i += 1;
if (nb != 0) {
buf_brelse(bp);
continue;
}
if (pref == 0)
pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0);
if (error =
ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, cred, &newb)) {
buf_brelse(bp);
goto fail;
}
nb = newb;
*allocblk++ = nb;
nbp = buf_getblk(vp, (daddr64_t)((unsigned)(indirs[i].in_lbn)), fs->fs_bsize, 0, 0, BLK_META);
buf_setblkno(nbp, (daddr64_t)((unsigned)fsbtodb(fs, nb)));
buf_clear(nbp);
/*
* Write synchronously conditional on mount flags.
*/
if ((vp)->v_mount->mnt_flag & MNT_ASYNC) {
error = 0;
buf_bdwrite(nbp);
} else if (error = buf_bwrite(nbp)) {
buf_brelse(bp);
goto fail;
}
#if REV_ENDIAN_FS
if (rev_endian)
bap[indirs[i - 1].in_off] = OSSwapInt32(nb);
else {
#endif /* REV_ENDIAN_FS */
bap[indirs[i - 1].in_off] = nb;
#if REV_ENDIAN_FS
}
#endif /* REV_ENDIAN_FS */
/*
* If required, write synchronously, otherwise use
* delayed write.
*/
if (flags & B_SYNC) {
buf_bwrite(bp);
} else {
buf_bdwrite(bp);
}
}
/*
* Get the data block, allocating if necessary.
*/
if (nb == 0) {
pref = ffs_blkpref(ip, lbn, indirs[i].in_off, &bap[0]);
if (error = ffs_alloc(ip,
lbn, pref, (int)fs->fs_bsize, cred, &newb)) {
buf_brelse(bp);
goto fail;
}
nb = newb;
*allocblk++ = nb;
#if REV_ENDIAN_FS
if (rev_endian)
bap[indirs[i].in_off] = OSSwapInt32(nb);
else {
#endif /* REV_ENDIAN_FS */
bap[indirs[i].in_off] = nb;
#if REV_ENDIAN_FS
}
#endif /* REV_ENDIAN_FS */
/*
* If required, write synchronously, otherwise use
* delayed write.
*/
if (flags & B_SYNC) {
buf_bwrite(bp);
} else {
buf_bdwrite(bp);
}
return (0);
}
buf_brelse(bp);
return (0);
fail:
/*
* If we have failed part way through block allocation, we
* have to deallocate any indirect blocks that we have allocated.
*/
for (deallocated = 0, blkp = allociblk; blkp < allocblk; blkp++) {
ffs_blkfree(ip, *blkp, fs->fs_bsize);
deallocated += fs->fs_bsize;
}
if (allocib != NULL)
*allocib = 0;
if (deallocated) {
devBlockSize = vfs_devblocksize(mp);
#if QUOTA
/*
* Restore user's disk quota because allocation failed.
*/
(void) chkdq(ip, (int64_t)-deallocated, cred, FORCE);
#endif /* QUOTA */
ip->i_blocks -= btodb(deallocated, devBlockSize);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
}
return (error);
}
int
ckinode(union dinode *dp, struct inodesc *idesc)
{
off_t remsize, sizepb;
int i, offset, ret;
union dinode dino;
ufs2_daddr_t ndb;
mode_t mode;
char pathbuf[MAXPATHLEN + 1];
if (idesc->id_fix != IGNORE)
idesc->id_fix = DONTKNOW;
idesc->id_lbn = -1;
idesc->id_entryno = 0;
idesc->id_filesize = DIP(dp, di_size);
mode = DIP(dp, di_mode) & IFMT;
if (mode == IFBLK || mode == IFCHR || (mode == IFLNK &&
DIP(dp, di_size) < (unsigned)sblock.fs_maxsymlinklen))
return (KEEPON);
if (sblock.fs_magic == FS_UFS1_MAGIC)
dino.dp1 = dp->dp1;
else
dino.dp2 = dp->dp2;
ndb = howmany(DIP(&dino, di_size), sblock.fs_bsize);
for (i = 0; i < NDADDR; i++) {
idesc->id_lbn++;
if (--ndb == 0 &&
(offset = blkoff(&sblock, DIP(&dino, di_size))) != 0)
idesc->id_numfrags =
numfrags(&sblock, fragroundup(&sblock, offset));
else
idesc->id_numfrags = sblock.fs_frag;
if (DIP(&dino, di_db[i]) == 0) {
if (idesc->id_type == DATA && ndb >= 0) {
/* An empty block in a directory XXX */
getpathname(pathbuf, idesc->id_number,
idesc->id_number);
pfatal("DIRECTORY %s: CONTAINS EMPTY BLOCKS",
pathbuf);
if (reply("ADJUST LENGTH") == 1) {
dp = ginode(idesc->id_number);
DIP_SET(dp, di_size,
i * sblock.fs_bsize);
printf(
"YOU MUST RERUN FSCK AFTERWARDS\n");
rerun = 1;
inodirty();
}
}
continue;
}
idesc->id_blkno = DIP(&dino, di_db[i]);
if (idesc->id_type != DATA)
ret = (*idesc->id_func)(idesc);
else
ret = dirscan(idesc);
if (ret & STOP)
return (ret);
}
idesc->id_numfrags = sblock.fs_frag;
remsize = DIP(&dino, di_size) - sblock.fs_bsize * NDADDR;
sizepb = sblock.fs_bsize;
for (i = 0; i < NIADDR; i++) {
sizepb *= NINDIR(&sblock);
if (DIP(&dino, di_ib[i])) {
idesc->id_blkno = DIP(&dino, di_ib[i]);
ret = iblock(idesc, i + 1, remsize, BT_LEVEL1 + i);
if (ret & STOP)
return (ret);
} else {
idesc->id_lbn += sizepb / sblock.fs_bsize;
if (idesc->id_type == DATA && remsize > 0) {
/* An empty block in a directory XXX */
getpathname(pathbuf, idesc->id_number,
idesc->id_number);
pfatal("DIRECTORY %s: CONTAINS EMPTY BLOCKS",
pathbuf);
if (reply("ADJUST LENGTH") == 1) {
dp = ginode(idesc->id_number);
DIP_SET(dp, di_size,
DIP(dp, di_size) - remsize);
remsize = 0;
printf(
"YOU MUST RERUN FSCK AFTERWARDS\n");
rerun = 1;
inodirty();
break;
}
}
}
remsize -= sizepb;
}
return (KEEPON);
}
static int ufs_mkdir(uufsd_t *ufs, ino_t parent, ino_t inum, char *name)
{
int retval;
struct ufs_vnode *parent_vnode = NULL, *vnode = NULL;
struct inode *parent_inode, *inode;
ino_t ino = inum;
ino_t scratch_ino;
ufs2_daddr_t blk;
char *block = 0;
struct fs *fs = &ufs->d_fs;
int dirsize = DIRBLKSIZ;
int blocksize = fragroundup(fs, dirsize);
parent_vnode = vnode_get(ufs, parent);
if (!parent_vnode) {
return ENOENT;
}
parent_inode = vnode2inode(parent_vnode);
/*
* Allocate an inode, if necessary
*/
if (!ino) {
retval = ufs_valloc(parent_vnode, DTTOIF(DT_DIR), &vnode);
if (retval)
goto cleanup;
ino = vnode->inode.i_number;
inode = vnode2inode(vnode);
}
/*
* Allocate a data block for the directory
*/
retval = ufs_block_alloc(ufs, inode, fragroundup(fs, dirsize), &blk);
if (retval)
goto cleanup;
/*
* Create a scratch template for the directory
*/
retval = ufs_new_dir_block(ufs, vnode->inode.i_number, parent_vnode, &block);
if (retval)
goto cleanup;
/*
* Get the parent's inode, if necessary
if (parent != ino) {
parent_vnode = vnode_get(ufs, parent);
if (retval)
goto cleanup;
} else
memset(&parent_inode, 0, sizeof(parent_inode));
*/
/*
* Create the inode structure....
*/
inode->i_mode = DT_DIR | (0777);
inode->i_uid = inode->i_gid = 0;
UFS_DINODE(inode)->di_db[0] = blk;
inode->i_nlink = 1;
inode->i_size = dirsize;
/*
* Write out the inode and inode data block
*/
retval = blkwrite(ufs, fsbtodb(fs, blk), block, blocksize);
if (retval == -1)
goto cleanup;
/*
* Link the directory into the filesystem hierarchy
*/
if (name) {
retval = ufs_lookup(ufs, parent, name, strlen(name),
&scratch_ino);
if (!retval) {
retval = EEXIST;
name = 0;
goto cleanup;
}
if (retval != ENOENT)
goto cleanup;
retval = ufs_link(ufs, parent, name, vnode, DTTOIF(DT_DIR));
if (retval)
goto cleanup;
}
/*
* Update parent inode's counts
*/
if (parent != ino) {
parent_inode->i_nlink++;
}
cleanup:
if (vnode)
vnode_put(vnode, 1);
if (parent_vnode)
vnode_put(parent_vnode, 1);
if (block)
ufs_free_mem(&block);
return retval;
}
/*
* Balloc defines the structure of filesystem storage
* by allocating the physical blocks on a device given
* the inode and the logical block number in a file.
* This is the allocation strategy for UFS1. Below is
* the allocation strategy for UFS2.
*/
int
ffs_balloc_ufs1(struct vnode *vp, off_t startoffset, int size,
struct ucred *cred, int flags, struct buf **bpp)
{
struct inode *ip;
struct ufs1_dinode *dp;
ufs_lbn_t lbn, lastlbn;
struct fs *fs;
ufs1_daddr_t nb;
struct buf *bp, *nbp;
struct ufsmount *ump;
struct indir indirs[NIADDR + 2];
int deallocated, osize, nsize, num, i, error;
ufs2_daddr_t newb;
ufs1_daddr_t *bap, pref;
ufs1_daddr_t *allocib, *blkp, *allocblk, allociblk[NIADDR + 1];
ufs2_daddr_t *lbns_remfree, lbns[NIADDR + 1];
int unwindidx = -1;
int saved_inbdflush;
static struct timeval lastfail;
static int curfail;
int reclaimed;
ip = VTOI(vp);
dp = ip->i_din1;
fs = ip->i_fs;
ump = ip->i_ump;
lbn = lblkno(fs, startoffset);
size = blkoff(fs, startoffset) + size;
reclaimed = 0;
if (size > fs->fs_bsize)
panic("ffs_balloc_ufs1: blk too big");
*bpp = NULL;
if (flags & IO_EXT)
return (EOPNOTSUPP);
if (lbn < 0)
return (EFBIG);
if (DOINGSOFTDEP(vp))
softdep_prealloc(vp, MNT_WAIT);
/*
* If the next write will extend the file into a new block,
* and the file is currently composed of a fragment
* this fragment has to be extended to be a full block.
*/
lastlbn = lblkno(fs, ip->i_size);
if (lastlbn < NDADDR && lastlbn < lbn) {
nb = lastlbn;
osize = blksize(fs, ip, nb);
if (osize < fs->fs_bsize && osize > 0) {
UFS_LOCK(ump);
error = ffs_realloccg(ip, nb, dp->di_db[nb],
ffs_blkpref_ufs1(ip, lastlbn, (int)nb,
&dp->di_db[0]), osize, (int)fs->fs_bsize, flags,
cred, &bp);
if (error)
return (error);
if (DOINGSOFTDEP(vp))
softdep_setup_allocdirect(ip, nb,
dbtofsb(fs, bp->b_blkno), dp->di_db[nb],
fs->fs_bsize, osize, bp);
ip->i_size = smalllblktosize(fs, nb + 1);
dp->di_size = ip->i_size;
dp->di_db[nb] = dbtofsb(fs, bp->b_blkno);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
if (flags & IO_SYNC)
bwrite(bp);
else
bawrite(bp);
}
}
/*
* The first NDADDR blocks are direct blocks
*/
if (lbn < NDADDR) {
if (flags & BA_METAONLY)
panic("ffs_balloc_ufs1: BA_METAONLY for direct block");
nb = dp->di_db[lbn];
if (nb != 0 && ip->i_size >= smalllblktosize(fs, lbn + 1)) {
error = bread(vp, lbn, fs->fs_bsize, NOCRED, &bp);
if (error) {
brelse(bp);
return (error);
}
bp->b_blkno = fsbtodb(fs, nb);
*bpp = bp;
return (0);
}
if (nb != 0) {
/*
* Consider need to reallocate a fragment.
*/
osize = fragroundup(fs, blkoff(fs, ip->i_size));
nsize = fragroundup(fs, size);
if (nsize <= osize) {
error = bread(vp, lbn, osize, NOCRED, &bp);
if (error) {
brelse(bp);
return (error);
}
bp->b_blkno = fsbtodb(fs, nb);
} else {
UFS_LOCK(ump);
error = ffs_realloccg(ip, lbn, dp->di_db[lbn],
ffs_blkpref_ufs1(ip, lbn, (int)lbn,
&dp->di_db[0]), osize, nsize, flags,
cred, &bp);
if (error)
return (error);
if (DOINGSOFTDEP(vp))
softdep_setup_allocdirect(ip, lbn,
dbtofsb(fs, bp->b_blkno), nb,
nsize, osize, bp);
}
} else {
if (ip->i_size < smalllblktosize(fs, lbn + 1))
nsize = fragroundup(fs, size);
else
nsize = fs->fs_bsize;
UFS_LOCK(ump);
error = ffs_alloc(ip, lbn,
ffs_blkpref_ufs1(ip, lbn, (int)lbn, &dp->di_db[0]),
nsize, flags, cred, &newb);
if (error)
return (error);
bp = getblk(vp, lbn, nsize, 0, 0, 0);
bp->b_blkno = fsbtodb(fs, newb);
if (flags & BA_CLRBUF)
vfs_bio_clrbuf(bp);
if (DOINGSOFTDEP(vp))
softdep_setup_allocdirect(ip, lbn, newb, 0,
nsize, 0, bp);
}
dp->di_db[lbn] = dbtofsb(fs, bp->b_blkno);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
*bpp = bp;
return (0);
}
/*
* Determine the number of levels of indirection.
*/
pref = 0;
if ((error = ufs_getlbns(vp, lbn, indirs, &num)) != 0)
return(error);
#ifdef INVARIANTS
if (num < 1)
panic ("ffs_balloc_ufs1: ufs_getlbns returned indirect block");
#endif
saved_inbdflush = curthread_pflags_set(TDP_INBDFLUSH);
/*
* Fetch the first indirect block allocating if necessary.
*/
--num;
nb = dp->di_ib[indirs[0].in_off];
allocib = NULL;
allocblk = allociblk;
lbns_remfree = lbns;
if (nb == 0) {
UFS_LOCK(ump);
pref = ffs_blkpref_ufs1(ip, lbn, 0, (ufs1_daddr_t *)0);
if ((error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize,
flags, cred, &newb)) != 0) {
curthread_pflags_restore(saved_inbdflush);
return (error);
}
nb = newb;
*allocblk++ = nb;
*lbns_remfree++ = indirs[1].in_lbn;
bp = getblk(vp, indirs[1].in_lbn, fs->fs_bsize, 0, 0, 0);
bp->b_blkno = fsbtodb(fs, nb);
vfs_bio_clrbuf(bp);
if (DOINGSOFTDEP(vp)) {
softdep_setup_allocdirect(ip, NDADDR + indirs[0].in_off,
newb, 0, fs->fs_bsize, 0, bp);
bdwrite(bp);
} else {
/*
* Write synchronously so that indirect blocks
* never point at garbage.
*/
if (DOINGASYNC(vp))
bdwrite(bp);
else if ((error = bwrite(bp)) != 0)
goto fail;
}
allocib = &dp->di_ib[indirs[0].in_off];
*allocib = nb;
ip->i_flag |= IN_CHANGE | IN_UPDATE;
}
/*
* Fetch through the indirect blocks, allocating as necessary.
*/
retry:
for (i = 1;;) {
error = bread(vp,
indirs[i].in_lbn, (int)fs->fs_bsize, NOCRED, &bp);
if (error) {
brelse(bp);
goto fail;
}
bap = (ufs1_daddr_t *)bp->b_data;
nb = bap[indirs[i].in_off];
if (i == num)
break;
i += 1;
if (nb != 0) {
bqrelse(bp);
continue;
}
UFS_LOCK(ump);
if (pref == 0)
pref = ffs_blkpref_ufs1(ip, lbn, 0, (ufs1_daddr_t *)0);
if ((error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize,
flags | IO_BUFLOCKED, cred, &newb)) != 0) {
brelse(bp);
if (++reclaimed == 1) {
UFS_LOCK(ump);
softdep_request_cleanup(fs, vp, cred,
FLUSH_BLOCKS_WAIT);
UFS_UNLOCK(ump);
goto retry;
}
if (ppsratecheck(&lastfail, &curfail, 1)) {
ffs_fserr(fs, ip->i_number, "filesystem full");
uprintf("\n%s: write failed, filesystem "
"is full\n", fs->fs_fsmnt);
}
goto fail;
}
nb = newb;
*allocblk++ = nb;
*lbns_remfree++ = indirs[i].in_lbn;
nbp = getblk(vp, indirs[i].in_lbn, fs->fs_bsize, 0, 0, 0);
nbp->b_blkno = fsbtodb(fs, nb);
vfs_bio_clrbuf(nbp);
if (DOINGSOFTDEP(vp)) {
softdep_setup_allocindir_meta(nbp, ip, bp,
indirs[i - 1].in_off, nb);
bdwrite(nbp);
} else {
/*
* Write synchronously so that indirect blocks
* never point at garbage.
*/
if ((error = bwrite(nbp)) != 0) {
brelse(bp);
goto fail;
}
}
bap[indirs[i - 1].in_off] = nb;
if (allocib == NULL && unwindidx < 0)
unwindidx = i - 1;
/*
* If required, write synchronously, otherwise use
* delayed write.
*/
if (flags & IO_SYNC) {
bwrite(bp);
} else {
if (bp->b_bufsize == fs->fs_bsize)
bp->b_flags |= B_CLUSTEROK;
bdwrite(bp);
}
}
/*
* If asked only for the indirect block, then return it.
*/
if (flags & BA_METAONLY) {
curthread_pflags_restore(saved_inbdflush);
*bpp = bp;
return (0);
}
/*
* Get the data block, allocating if necessary.
*/
if (nb == 0) {
UFS_LOCK(ump);
pref = ffs_blkpref_ufs1(ip, lbn, indirs[i].in_off, &bap[0]);
error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize,
flags | IO_BUFLOCKED, cred, &newb);
if (error) {
brelse(bp);
if (++reclaimed == 1) {
UFS_LOCK(ump);
softdep_request_cleanup(fs, vp, cred,
FLUSH_BLOCKS_WAIT);
UFS_UNLOCK(ump);
goto retry;
}
if (ppsratecheck(&lastfail, &curfail, 1)) {
ffs_fserr(fs, ip->i_number, "filesystem full");
uprintf("\n%s: write failed, filesystem "
"is full\n", fs->fs_fsmnt);
}
goto fail;
}
nb = newb;
*allocblk++ = nb;
*lbns_remfree++ = lbn;
nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0, 0);
nbp->b_blkno = fsbtodb(fs, nb);
if (flags & BA_CLRBUF)
vfs_bio_clrbuf(nbp);
if (DOINGSOFTDEP(vp))
softdep_setup_allocindir_page(ip, lbn, bp,
indirs[i].in_off, nb, 0, nbp);
bap[indirs[i].in_off] = nb;
/*
* If required, write synchronously, otherwise use
* delayed write.
*/
if (flags & IO_SYNC) {
bwrite(bp);
} else {
if (bp->b_bufsize == fs->fs_bsize)
bp->b_flags |= B_CLUSTEROK;
bdwrite(bp);
}
curthread_pflags_restore(saved_inbdflush);
*bpp = nbp;
return (0);
}
brelse(bp);
if (flags & BA_CLRBUF) {
int seqcount = (flags & BA_SEQMASK) >> BA_SEQSHIFT;
if (seqcount && (vp->v_mount->mnt_flag & MNT_NOCLUSTERR) == 0) {
error = cluster_read(vp, ip->i_size, lbn,
(int)fs->fs_bsize, NOCRED,
MAXBSIZE, seqcount, &nbp);
} else {
error = bread(vp, lbn, (int)fs->fs_bsize, NOCRED, &nbp);
}
if (error) {
brelse(nbp);
goto fail;
}
} else {
int
setup(char *dev)
{
long cg, size, asked, i, j, bmapsize;
struct disklabel *lp;
off_t sizepb;
struct stat statb;
struct fs proto;
int doskipclean;
int32_t maxsymlinklen, nindir, inopb;
u_int64_t maxfilesize;
char *realdev;
havesb = 0;
fswritefd = fsreadfd = -1;
doskipclean = skipclean;
if ((fsreadfd = opendev(dev, O_RDONLY, 0, &realdev)) < 0) {
printf("Can't open %s: %s\n", dev, strerror(errno));
return (0);
}
if (strncmp(dev, realdev, PATH_MAX) != 0) {
blockcheck(unrawname(realdev));
strlcpy(rdevname, realdev, sizeof(rdevname));
setcdevname(rdevname, dev, preen);
}
if (fstat(fsreadfd, &statb) < 0) {
printf("Can't stat %s: %s\n", realdev, strerror(errno));
close(fsreadfd);
return (0);
}
if (!S_ISCHR(statb.st_mode)) {
pfatal("%s is not a character device", realdev);
if (reply("CONTINUE") == 0) {
close(fsreadfd);
return (0);
}
}
if (preen == 0) {
printf("** %s", realdev);
if (strncmp(dev, realdev, PATH_MAX) != 0)
printf(" (%s)", dev);
}
if (nflag || (fswritefd = opendev(dev, O_WRONLY, 0, NULL)) < 0) {
fswritefd = -1;
if (preen)
pfatal("NO WRITE ACCESS");
printf(" (NO WRITE)");
}
if (preen == 0)
printf("\n");
fsmodified = 0;
lfdir = 0;
initbarea(&sblk);
initbarea(&asblk);
sblk.b_un.b_buf = malloc(SBSIZE);
asblk.b_un.b_buf = malloc(SBSIZE);
if (sblk.b_un.b_buf == NULL || asblk.b_un.b_buf == NULL)
errexit("cannot allocate space for superblock\n");
if ((lp = getdisklabel(NULL, fsreadfd)) != NULL)
dev_bsize = secsize = lp->d_secsize;
else
dev_bsize = secsize = DEV_BSIZE;
/*
* Read in the superblock, looking for alternates if necessary
*/
if (readsb(1) == 0) {
if (bflag || preen || calcsb(realdev, fsreadfd, &proto) == 0)
return(0);
if (reply("LOOK FOR ALTERNATE SUPERBLOCKS") == 0)
return (0);
for (i = 0; i < sizeof(altsbtry) / sizeof(altsbtry[0]); i++) {
bflag = altsbtry[i];
/* proto partially setup by calcsb */
if (readsb(0) != 0 &&
proto.fs_fsize == sblock.fs_fsize &&
proto.fs_bsize == sblock.fs_bsize)
goto found;
}
for (cg = 0; cg < proto.fs_ncg; cg++) {
bflag = fsbtodb(&proto, cgsblock(&proto, cg));
if (readsb(0) != 0)
break;
}
if (cg >= proto.fs_ncg) {
printf("%s %s\n%s %s\n%s %s\n",
"SEARCH FOR ALTERNATE SUPER-BLOCK",
"FAILED. YOU MUST USE THE",
"-b OPTION TO FSCK_FFS TO SPECIFY THE",
"LOCATION OF AN ALTERNATE",
"SUPER-BLOCK TO SUPPLY NEEDED",
"INFORMATION; SEE fsck_ffs(8).");
return(0);
}
found:
doskipclean = 0;
pwarn("USING ALTERNATE SUPERBLOCK AT %d\n", bflag);
}
if (debug)
printf("clean = %d\n", sblock.fs_clean);
if (sblock.fs_clean & FS_ISCLEAN) {
if (doskipclean) {
pwarn("%sile system is clean; not checking\n",
preen ? "f" : "** F");
return (-1);
}
if (!preen)
pwarn("** File system is already clean\n");
}
maxfsblock = sblock.fs_size;
maxino = sblock.fs_ncg * sblock.fs_ipg;
sizepb = sblock.fs_bsize;
maxfilesize = sblock.fs_bsize * NDADDR - 1;
for (i = 0; i < NIADDR; i++) {
sizepb *= NINDIR(&sblock);
maxfilesize += sizepb;
}
/*
* Check and potentially fix certain fields in the super block.
*/
if (sblock.fs_optim != FS_OPTTIME && sblock.fs_optim != FS_OPTSPACE) {
pfatal("UNDEFINED OPTIMIZATION IN SUPERBLOCK");
if (reply("SET TO DEFAULT") == 1) {
sblock.fs_optim = FS_OPTTIME;
sbdirty();
}
}
if ((sblock.fs_minfree < 0 || sblock.fs_minfree > 99)) {
pfatal("IMPOSSIBLE MINFREE=%d IN SUPERBLOCK",
sblock.fs_minfree);
if (reply("SET TO DEFAULT") == 1) {
sblock.fs_minfree = 10;
sbdirty();
}
}
if (sblock.fs_npsect < sblock.fs_nsect ||
sblock.fs_npsect > sblock.fs_nsect*2) {
pwarn("IMPOSSIBLE NPSECT=%d IN SUPERBLOCK",
sblock.fs_npsect);
sblock.fs_npsect = sblock.fs_nsect;
if (preen)
printf(" (FIXED)\n");
if (preen || reply("SET TO DEFAULT") == 1) {
sbdirty();
dirty(&asblk);
}
}
if (sblock.fs_bmask != ~(sblock.fs_bsize - 1)) {
pwarn("INCORRECT BMASK=%x IN SUPERBLOCK",
sblock.fs_bmask);
sblock.fs_bmask = ~(sblock.fs_bsize - 1);
if (preen)
printf(" (FIXED)\n");
if (preen || reply("FIX") == 1) {
sbdirty();
dirty(&asblk);
}
}
if (sblock.fs_fmask != ~(sblock.fs_fsize - 1)) {
pwarn("INCORRECT FMASK=%x IN SUPERBLOCK",
sblock.fs_fmask);
sblock.fs_fmask = ~(sblock.fs_fsize - 1);
if (preen)
printf(" (FIXED)\n");
if (preen || reply("FIX") == 1) {
sbdirty();
dirty(&asblk);
}
}
if (1 << sblock.fs_bshift != sblock.fs_bsize) {
pwarn("INCORRECT BSHIFT=%d IN SUPERBLOCK", sblock.fs_bshift);
sblock.fs_bshift = ffs(sblock.fs_bsize) - 1;
if (preen)
printf(" (FIXED)\n");
if (preen || reply("FIX") == 1) {
sbdirty();
dirty(&asblk);
}
}
if (1 << sblock.fs_fshift != sblock.fs_fsize) {
pwarn("INCORRECT FSHIFT=%d IN SUPERBLOCK", sblock.fs_fshift);
sblock.fs_fshift = ffs(sblock.fs_fsize) - 1;
if (preen)
printf(" (FIXED)\n");
if (preen || reply("FIX") == 1) {
sbdirty();
dirty(&asblk);
}
}
if (sblock.fs_inodefmt < FS_44INODEFMT) {
pwarn("Format of filesystem is too old.\n");
pwarn("Must update to modern format using a version of fsck\n");
pfatal("from before release 5.0 with the command ``fsck -c 2''\n");
exit(8);
}
if (sblock.fs_maxfilesize != maxfilesize) {
pwarn("INCORRECT MAXFILESIZE=%llu IN SUPERBLOCK",
(unsigned long long)sblock.fs_maxfilesize);
sblock.fs_maxfilesize = maxfilesize;
if (preen)
printf(" (FIXED)\n");
if (preen || reply("FIX") == 1) {
sbdirty();
dirty(&asblk);
}
}
maxsymlinklen = sblock.fs_magic == FS_UFS1_MAGIC ?
MAXSYMLINKLEN_UFS1 : MAXSYMLINKLEN_UFS2;
if (sblock.fs_maxsymlinklen != maxsymlinklen) {
pwarn("INCORRECT MAXSYMLINKLEN=%d IN SUPERBLOCK",
sblock.fs_maxsymlinklen);
sblock.fs_maxsymlinklen = maxsymlinklen;
if (preen)
printf(" (FIXED)\n");
if (preen || reply("FIX") == 1) {
sbdirty();
dirty(&asblk);
}
}
if (sblock.fs_qbmask != ~sblock.fs_bmask) {
pwarn("INCORRECT QBMASK=%lx IN SUPERBLOCK",
(unsigned long)sblock.fs_qbmask);
sblock.fs_qbmask = ~sblock.fs_bmask;
if (preen)
printf(" (FIXED)\n");
if (preen || reply("FIX") == 1) {
sbdirty();
dirty(&asblk);
}
}
if (sblock.fs_qfmask != ~sblock.fs_fmask) {
pwarn("INCORRECT QFMASK=%lx IN SUPERBLOCK",
(unsigned long)sblock.fs_qfmask);
sblock.fs_qfmask = ~sblock.fs_fmask;
if (preen)
printf(" (FIXED)\n");
if (preen || reply("FIX") == 1) {
sbdirty();
dirty(&asblk);
}
}
if (sblock.fs_cgsize != fragroundup(&sblock, CGSIZE(&sblock))) {
pwarn("INCONSISTENT CGSIZE=%d\n", sblock.fs_cgsize);
sblock.fs_cgsize = fragroundup(&sblock, CGSIZE(&sblock));
if (preen)
printf(" (FIXED)\n");
if (preen || reply("FIX") == 1) {
sbdirty();
dirty(&asblk);
}
}
if (sblock.fs_magic == FS_UFS2_MAGIC)
inopb = sblock.fs_bsize / sizeof(struct ufs2_dinode);
else
inopb = sblock.fs_bsize / sizeof(struct ufs1_dinode);
if (INOPB(&sblock) != inopb) {
pwarn("INCONSISTENT INOPB=%d\n", INOPB(&sblock));
sblock.fs_inopb = inopb;
if (preen)
printf(" (FIXED)\n");
if (preen || reply("FIX") == 1) {
sbdirty();
dirty(&asblk);
}
}
if (sblock.fs_magic == FS_UFS2_MAGIC)
nindir = sblock.fs_bsize / sizeof(int64_t);
else
nindir = sblock.fs_bsize / sizeof(int32_t);
if (NINDIR(&sblock) != nindir) {
pwarn("INCONSISTENT NINDIR=%d\n", NINDIR(&sblock));
sblock.fs_nindir = nindir;
if (preen)
printf(" (FIXED)\n");
if (preen || reply("FIX") == 1) {
sbdirty();
dirty(&asblk);
}
}
if (asblk.b_dirty && !bflag) {
memcpy(&altsblock, &sblock, (size_t)sblock.fs_sbsize);
flush(fswritefd, &asblk);
}
/*
* read in the summary info.
*/
asked = 0;
sblock.fs_csp = calloc(1, sblock.fs_cssize);
if (sblock.fs_csp == NULL) {
printf("cannot alloc %u bytes for cylinder group summary area\n",
(unsigned)sblock.fs_cssize);
goto badsblabel;
}
for (i = 0, j = 0; i < sblock.fs_cssize; i += sblock.fs_bsize, j++) {
size = sblock.fs_cssize - i < sblock.fs_bsize ?
sblock.fs_cssize - i : sblock.fs_bsize;
if (bread(fsreadfd, (char *)sblock.fs_csp + i,
fsbtodb(&sblock, sblock.fs_csaddr + j * sblock.fs_frag),
size) != 0 && !asked) {
pfatal("BAD SUMMARY INFORMATION");
if (reply("CONTINUE") == 0) {
ckfini(0);
errexit("%s", "");
}
asked++;
}
}
/*
* allocate and initialize the necessary maps
*/
bmapsize = roundup(howmany(maxfsblock, NBBY), sizeof(int16_t));
blockmap = calloc((unsigned)bmapsize, sizeof(char));
if (blockmap == NULL) {
printf("cannot alloc %u bytes for blockmap\n",
(unsigned)bmapsize);
goto badsblabel;
}
inostathead = calloc((unsigned)(sblock.fs_ncg),
sizeof(struct inostatlist));
if (inostathead == NULL) {
printf("cannot alloc %u bytes for inostathead\n",
(unsigned)(sizeof(struct inostatlist) * (sblock.fs_ncg)));
goto badsblabel;
}
numdirs = MAX(sblock.fs_cstotal.cs_ndir, 128);
inplast = 0;
listmax = numdirs + 10;
inpsort = calloc((unsigned)listmax, sizeof(struct inoinfo *));
if (inpsort == NULL) {
printf("cannot alloc %zu bytes for inpsort\n",
(unsigned)listmax * sizeof(struct inoinfo *));
goto badsblabel;
}
inphead = calloc((unsigned)numdirs, sizeof(struct inoinfo *));
if (inphead == NULL) {
printf("cannot alloc %zu bytes for inphead\n",
(unsigned)numdirs * sizeof(struct inoinfo *));
goto badsblabel;
}
bufinit();
if (sblock.fs_flags & FS_DOSOFTDEP)
usedsoftdep = 1;
else
usedsoftdep = 0;
return (1);
badsblabel:
ckfini(0);
return (0);
}
/*
* Balloc defines the structure of file system storage
* by allocating the physical blocks on a device given
* the inode and the logical block number in a file.
*/
ffs_balloc(
register struct inode *ip,
register ufs_daddr_t lbn,
int size,
kauth_cred_t cred,
struct buf **bpp,
int flags,
int * blk_alloc)
{
register struct fs *fs;
register ufs_daddr_t nb;
struct buf *bp, *nbp;
struct vnode *vp = ITOV(ip);
struct indir indirs[NIADDR + 2];
ufs_daddr_t newb, *bap, pref;
int deallocated, osize, nsize, num, i, error;
ufs_daddr_t *allocib, *blkp, *allocblk, allociblk[NIADDR + 1];
int devBlockSize=0;
int alloc_buffer = 1;
struct mount *mp=vp->v_mount;
#if REV_ENDIAN_FS
int rev_endian=(mp->mnt_flag & MNT_REVEND);
#endif /* REV_ENDIAN_FS */
*bpp = NULL;
if (lbn < 0)
return (EFBIG);
fs = ip->i_fs;
if (flags & B_NOBUFF)
alloc_buffer = 0;
if (blk_alloc)
*blk_alloc = 0;
/*
* If the next write will extend the file into a new block,
* and the file is currently composed of a fragment
* this fragment has to be extended to be a full block.
*/
nb = lblkno(fs, ip->i_size);
if (nb < NDADDR && nb < lbn) {
/* the filesize prior to this write can fit in direct
* blocks (ie. fragmentaion is possibly done)
* we are now extending the file write beyond
* the block which has end of file prior to this write
*/
osize = blksize(fs, ip, nb);
/* osize gives disk allocated size in the last block. It is
* either in fragments or a file system block size */
if (osize < fs->fs_bsize && osize > 0) {
/* few fragments are already allocated,since the
* current extends beyond this block
* allocate the complete block as fragments are only
* in last block
*/
error = ffs_realloccg(ip, nb,
ffs_blkpref(ip, nb, (int)nb, &ip->i_db[0]),
osize, (int)fs->fs_bsize, cred, &bp);
if (error)
return (error);
/* adjust the inode size we just grew */
/* it is in nb+1 as nb starts from 0 */
ip->i_size = (nb + 1) * fs->fs_bsize;
ubc_setsize(vp, (off_t)ip->i_size);
ip->i_db[nb] = dbtofsb(fs, (ufs_daddr_t)buf_blkno(bp));
ip->i_flag |= IN_CHANGE | IN_UPDATE;
if ((flags & B_SYNC) || (!alloc_buffer)) {
if (!alloc_buffer)
buf_setflags(bp, B_NOCACHE);
buf_bwrite(bp);
} else
buf_bdwrite(bp);
/* note that bp is already released here */
}
}
/*
* The first NDADDR blocks are direct blocks
*/
if (lbn < NDADDR) {
nb = ip->i_db[lbn];
if (nb != 0 && ip->i_size >= (lbn + 1) * fs->fs_bsize) {
if (alloc_buffer) {
error = (int)buf_bread(vp, (daddr64_t)((unsigned)lbn), fs->fs_bsize, NOCRED, &bp);
if (error) {
buf_brelse(bp);
return (error);
}
*bpp = bp;
}
return (0);
}
if (nb != 0) {
/*
* Consider need to reallocate a fragment.
*/
osize = fragroundup(fs, blkoff(fs, ip->i_size));
nsize = fragroundup(fs, size);
if (nsize <= osize) {
if (alloc_buffer) {
error = (int)buf_bread(vp, (daddr64_t)((unsigned)lbn), osize, NOCRED, &bp);
if (error) {
buf_brelse(bp);
return (error);
}
ip->i_flag |= IN_CHANGE | IN_UPDATE;
*bpp = bp;
return (0);
}
else {
ip->i_flag |= IN_CHANGE | IN_UPDATE;
return (0);
}
} else {
error = ffs_realloccg(ip, lbn,
ffs_blkpref(ip, lbn, (int)lbn,
&ip->i_db[0]), osize, nsize, cred, &bp);
if (error)
return (error);
ip->i_db[lbn] = dbtofsb(fs, (ufs_daddr_t)buf_blkno(bp));
ip->i_flag |= IN_CHANGE | IN_UPDATE;
/* adjust the inode size we just grew */
ip->i_size = (lbn * fs->fs_bsize) + size;
ubc_setsize(vp, (off_t)ip->i_size);
if (!alloc_buffer) {
buf_setflags(bp, B_NOCACHE);
if (flags & B_SYNC)
buf_bwrite(bp);
else
buf_bdwrite(bp);
} else
*bpp = bp;
return (0);
}
} else {
if (ip->i_size < (lbn + 1) * fs->fs_bsize)
nsize = fragroundup(fs, size);
else
nsize = fs->fs_bsize;
error = ffs_alloc(ip, lbn,
ffs_blkpref(ip, lbn, (int)lbn, &ip->i_db[0]),
nsize, cred, &newb);
if (error)
return (error);
if (alloc_buffer) {
bp = buf_getblk(vp, (daddr64_t)((unsigned)lbn), nsize, 0, 0, BLK_WRITE);
buf_setblkno(bp, (daddr64_t)((unsigned)fsbtodb(fs, newb)));
if (flags & B_CLRBUF)
buf_clear(bp);
}
ip->i_db[lbn] = newb;
ip->i_flag |= IN_CHANGE | IN_UPDATE;
if (blk_alloc) {
*blk_alloc = nsize;
}
if (alloc_buffer)
*bpp = bp;
return (0);
}
}
/*
* Determine the number of levels of indirection.
*/
pref = 0;
if (error = ufs_getlbns(vp, lbn, indirs, &num))
return(error);
#if DIAGNOSTIC
if (num < 1)
panic ("ffs_balloc: ufs_bmaparray returned indirect block");
#endif
/*
* Fetch the first indirect block allocating if necessary.
*/
--num;
nb = ip->i_ib[indirs[0].in_off];
allocib = NULL;
allocblk = allociblk;
if (nb == 0) {
pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0);
if (error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize,
cred, &newb))
return (error);
nb = newb;
*allocblk++ = nb;
bp = buf_getblk(vp, (daddr64_t)((unsigned)(indirs[1].in_lbn)), fs->fs_bsize, 0, 0, BLK_META);
buf_setblkno(bp, (daddr64_t)((unsigned)fsbtodb(fs, nb)));
buf_clear(bp);
/*
* Write synchronously conditional on mount flags.
*/
if ((vp)->v_mount->mnt_flag & MNT_ASYNC) {
error = 0;
buf_bdwrite(bp);
} else if ((error = buf_bwrite(bp)) != 0) {
goto fail;
}
allocib = &ip->i_ib[indirs[0].in_off];
*allocib = nb;
ip->i_flag |= IN_CHANGE | IN_UPDATE;
}
/*
* Fetch through the indirect blocks, allocating as necessary.
*/
for (i = 1;;) {
error = (int)buf_meta_bread(vp, (daddr64_t)((unsigned)(indirs[i].in_lbn)), (int)fs->fs_bsize, NOCRED, &bp);
if (error) {
buf_brelse(bp);
goto fail;
}
bap = (ufs_daddr_t *)buf_dataptr(bp);
#if REV_ENDIAN_FS
if (rev_endian)
nb = OSSwapInt32(bap[indirs[i].in_off]);
else {
#endif /* REV_ENDIAN_FS */
nb = bap[indirs[i].in_off];
#if REV_ENDIAN_FS
}
#endif /* REV_ENDIAN_FS */
if (i == num)
break;
i += 1;
if (nb != 0) {
buf_brelse(bp);
continue;
}
if (pref == 0)
pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0);
if (error =
ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, cred, &newb)) {
buf_brelse(bp);
goto fail;
}
nb = newb;
*allocblk++ = nb;
nbp = buf_getblk(vp, (daddr64_t)((unsigned)(indirs[i].in_lbn)), fs->fs_bsize, 0, 0, BLK_META);
buf_setblkno(nbp, (daddr64_t)((unsigned)fsbtodb(fs, nb)));
buf_clear(nbp);
/*
* Write synchronously conditional on mount flags.
*/
if ((vp)->v_mount->mnt_flag & MNT_ASYNC) {
error = 0;
buf_bdwrite(nbp);
} else if (error = buf_bwrite(nbp)) {
buf_brelse(bp);
goto fail;
}
#if REV_ENDIAN_FS
if (rev_endian)
bap[indirs[i - 1].in_off] = OSSwapInt32(nb);
else {
#endif /* REV_ENDIAN_FS */
bap[indirs[i - 1].in_off] = nb;
#if REV_ENDIAN_FS
}
#endif /* REV_ENDIAN_FS */
/*
* If required, write synchronously, otherwise use
* delayed write.
*/
if (flags & B_SYNC) {
buf_bwrite(bp);
} else {
buf_bdwrite(bp);
}
}
/*
* Get the data block, allocating if necessary.
*/
if (nb == 0) {
pref = ffs_blkpref(ip, lbn, indirs[i].in_off, &bap[0]);
if (error = ffs_alloc(ip,
lbn, pref, (int)fs->fs_bsize, cred, &newb)) {
buf_brelse(bp);
goto fail;
}
nb = newb;
*allocblk++ = nb;
#if REV_ENDIAN_FS
if (rev_endian)
bap[indirs[i].in_off] = OSSwapInt32(nb);
else {
#endif /* REV_ENDIAN_FS */
bap[indirs[i].in_off] = nb;
#if REV_ENDIAN_FS
}
#endif /* REV_ENDIAN_FS */
/*
* If required, write synchronously, otherwise use
* delayed write.
*/
if ((flags & B_SYNC)) {
buf_bwrite(bp);
} else {
buf_bdwrite(bp);
}
if(alloc_buffer ) {
nbp = buf_getblk(vp, (daddr64_t)((unsigned)lbn), fs->fs_bsize, 0, 0, BLK_WRITE);
buf_setblkno(nbp, (daddr64_t)((unsigned)fsbtodb(fs, nb)));
if (flags & B_CLRBUF)
buf_clear(nbp);
}
if (blk_alloc) {
*blk_alloc = fs->fs_bsize;
}
if(alloc_buffer)
*bpp = nbp;
return (0);
}
buf_brelse(bp);
if (alloc_buffer) {
if (flags & B_CLRBUF) {
error = (int)buf_bread(vp, (daddr64_t)((unsigned)lbn), (int)fs->fs_bsize, NOCRED, &nbp);
if (error) {
buf_brelse(nbp);
goto fail;
}
} else {
nbp = buf_getblk(vp, (daddr64_t)((unsigned)lbn), fs->fs_bsize, 0, 0, BLK_WRITE);
buf_setblkno(nbp, (daddr64_t)((unsigned)fsbtodb(fs, nb)));
}
*bpp = nbp;
}
return (0);
fail:
/*
* If we have failed part way through block allocation, we
* have to deallocate any indirect blocks that we have allocated.
*/
for (deallocated = 0, blkp = allociblk; blkp < allocblk; blkp++) {
ffs_blkfree(ip, *blkp, fs->fs_bsize);
deallocated += fs->fs_bsize;
}
if (allocib != NULL)
*allocib = 0;
if (deallocated) {
devBlockSize = vfs_devblocksize(mp);
#if QUOTA
/*
* Restore user's disk quota because allocation failed.
*/
(void) chkdq(ip, (int64_t)-deallocated, cred, FORCE);
#endif /* QUOTA */
ip->i_blocks -= btodb(deallocated, devBlockSize);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
}
return (error);
}
/*
* ffs_balloc(struct vnode *a_vp, ufs_daddr_t a_lbn, int a_size,
* struct ucred *a_cred, int a_flags, struct buf *a_bpp)
*
* Balloc defines the structure of filesystem storage by allocating
* the physical blocks on a device given the inode and the logical
* block number in a file.
*
* NOTE: B_CLRBUF - this flag tells balloc to clear invalid portions
* of the buffer. However, any dirty bits will override missing
* valid bits. This case occurs when writable mmaps are truncated
* and then extended.
*/
int
ffs_balloc(struct vop_balloc_args *ap)
{
struct inode *ip;
ufs_daddr_t lbn;
int size;
struct ucred *cred;
int flags;
struct fs *fs;
ufs_daddr_t nb;
struct buf *bp, *nbp, *dbp;
struct vnode *vp;
struct indir indirs[NIADDR + 2];
ufs_daddr_t newb, *bap, pref;
int deallocated, osize, nsize, num, i, error;
ufs_daddr_t *allocib, *blkp, *allocblk, allociblk[NIADDR + 1];
ufs_daddr_t *lbns_remfree, lbns[NIADDR + 1];
int unwindidx;
int seqcount;
vp = ap->a_vp;
ip = VTOI(vp);
fs = ip->i_fs;
lbn = lblkno(fs, ap->a_startoffset);
size = blkoff(fs, ap->a_startoffset) + ap->a_size;
if (size > fs->fs_bsize)
panic("ffs_balloc: blk too big");
*ap->a_bpp = NULL;
if (lbn < 0)
return (EFBIG);
cred = ap->a_cred;
flags = ap->a_flags;
/*
* The vnode must be locked for us to be able to safely mess
* around with the inode.
*/
if (vn_islocked(vp) != LK_EXCLUSIVE) {
panic("ffs_balloc: vnode %p not exclusively locked!", vp);
}
/*
* If the next write will extend the file into a new block,
* and the file is currently composed of a fragment
* this fragment has to be extended to be a full block.
*/
nb = lblkno(fs, ip->i_size);
if (nb < NDADDR && nb < lbn) {
/*
* The filesize prior to this write can fit in direct
* blocks (ex. fragmentation is possibly done)
* we are now extending the file write beyond
* the block which has end of the file prior to this write.
*/
osize = blksize(fs, ip, nb);
/*
* osize gives disk allocated size in the last block. It is
* either in fragments or a file system block size.
*/
if (osize < fs->fs_bsize && osize > 0) {
/* A few fragments are already allocated, since the
* current extends beyond this block allocated the
* complete block as fragments are on in last block.
*/
error = ffs_realloccg(ip, nb,
ffs_blkpref(ip, nb, (int)nb, &ip->i_db[0]),
osize, (int)fs->fs_bsize, cred, &bp);
if (error)
return (error);
if (DOINGSOFTDEP(vp))
softdep_setup_allocdirect(ip, nb,
dofftofsb(fs, bp->b_bio2.bio_offset),
ip->i_db[nb], fs->fs_bsize, osize, bp);
/* adjust the inode size, we just grew */
ip->i_size = smalllblktosize(fs, nb + 1);
ip->i_db[nb] = dofftofsb(fs, bp->b_bio2.bio_offset);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
if (flags & B_SYNC)
bwrite(bp);
else
bawrite(bp);
/* bp is already released here */
}
}
/*
* The first NDADDR blocks are direct blocks
*/
if (lbn < NDADDR) {
nb = ip->i_db[lbn];
if (nb != 0 && ip->i_size >= smalllblktosize(fs, lbn + 1)) {
error = bread(vp, lblktodoff(fs, lbn), fs->fs_bsize, &bp);
if (error) {
brelse(bp);
return (error);
}
bp->b_bio2.bio_offset = fsbtodoff(fs, nb);
*ap->a_bpp = bp;
return (0);
}
if (nb != 0) {
/*
* Consider need to reallocate a fragment.
*/
osize = fragroundup(fs, blkoff(fs, ip->i_size));
nsize = fragroundup(fs, size);
if (nsize <= osize) {
error = bread(vp, lblktodoff(fs, lbn),
osize, &bp);
if (error) {
brelse(bp);
return (error);
}
bp->b_bio2.bio_offset = fsbtodoff(fs, nb);
} else {
/*
* NOTE: ffs_realloccg() issues a bread().
*/
error = ffs_realloccg(ip, lbn,
ffs_blkpref(ip, lbn, (int)lbn,
&ip->i_db[0]), osize, nsize, cred, &bp);
if (error)
return (error);
if (DOINGSOFTDEP(vp))
softdep_setup_allocdirect(ip, lbn,
dofftofsb(fs, bp->b_bio2.bio_offset),
nb, nsize, osize, bp);
}
} else {
if (ip->i_size < smalllblktosize(fs, lbn + 1))
nsize = fragroundup(fs, size);
else
nsize = fs->fs_bsize;
error = ffs_alloc(ip, lbn,
ffs_blkpref(ip, lbn, (int)lbn, &ip->i_db[0]),
nsize, cred, &newb);
if (error)
return (error);
bp = getblk(vp, lblktodoff(fs, lbn), nsize, 0, 0);
bp->b_bio2.bio_offset = fsbtodoff(fs, newb);
if (flags & B_CLRBUF)
vfs_bio_clrbuf(bp);
if (DOINGSOFTDEP(vp))
softdep_setup_allocdirect(ip, lbn, newb, 0,
nsize, 0, bp);
}
ip->i_db[lbn] = dofftofsb(fs, bp->b_bio2.bio_offset);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
*ap->a_bpp = bp;
return (0);
}
/*
* Determine the number of levels of indirection.
*/
pref = 0;
if ((error = ufs_getlbns(vp, lbn, indirs, &num)) != 0)
return(error);
#ifdef DIAGNOSTIC
if (num < 1)
panic ("ffs_balloc: ufs_bmaparray returned indirect block");
#endif
/*
* Get a handle on the data block buffer before working through
* indirect blocks to avoid a deadlock between the VM system holding
* a locked VM page and issuing a BMAP (which tries to lock the
* indirect blocks), and the filesystem holding a locked indirect
* block and then trying to read a data block (which tries to lock
* the underlying VM pages).
*/
dbp = getblk(vp, lblktodoff(fs, lbn), fs->fs_bsize, 0, 0);
/*
* Setup undo history
*/
allocib = NULL;
allocblk = allociblk;
lbns_remfree = lbns;
unwindidx = -1;
/*
* Fetch the first indirect block directly from the inode, allocating
* one if necessary.
*/
--num;
nb = ip->i_ib[indirs[0].in_off];
if (nb == 0) {
pref = ffs_blkpref(ip, lbn, 0, NULL);
/*
* If the filesystem has run out of space we can skip the
* full fsync/undo of the main [fail] case since no undo
* history has been built yet. Hence the goto fail2.
*/
if ((error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize,
cred, &newb)) != 0)
goto fail2;
nb = newb;
*allocblk++ = nb;
*lbns_remfree++ = indirs[1].in_lbn;
bp = getblk(vp, lblktodoff(fs, indirs[1].in_lbn),
fs->fs_bsize, 0, 0);
bp->b_bio2.bio_offset = fsbtodoff(fs, nb);
vfs_bio_clrbuf(bp);
if (DOINGSOFTDEP(vp)) {
softdep_setup_allocdirect(ip, NDADDR + indirs[0].in_off,
newb, 0, fs->fs_bsize, 0, bp);
bdwrite(bp);
} else {
/*
* Write synchronously so that indirect blocks
* never point at garbage.
*/
if (DOINGASYNC(vp))
bdwrite(bp);
else if ((error = bwrite(bp)) != 0)
goto fail;
}
allocib = &ip->i_ib[indirs[0].in_off];
*allocib = nb;
ip->i_flag |= IN_CHANGE | IN_UPDATE;
}
/*
* Fetch through the indirect blocks, allocating as necessary.
*/
for (i = 1;;) {
error = bread(vp, lblktodoff(fs, indirs[i].in_lbn), (int)fs->fs_bsize, &bp);
if (error) {
brelse(bp);
goto fail;
}
bap = (ufs_daddr_t *)bp->b_data;
nb = bap[indirs[i].in_off];
if (i == num)
break;
i += 1;
if (nb != 0) {
bqrelse(bp);
continue;
}
if (pref == 0)
pref = ffs_blkpref(ip, lbn, 0, NULL);
if ((error =
ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, cred, &newb)) != 0) {
brelse(bp);
goto fail;
}
nb = newb;
*allocblk++ = nb;
*lbns_remfree++ = indirs[i].in_lbn;
nbp = getblk(vp, lblktodoff(fs, indirs[i].in_lbn),
fs->fs_bsize, 0, 0);
nbp->b_bio2.bio_offset = fsbtodoff(fs, nb);
vfs_bio_clrbuf(nbp);
if (DOINGSOFTDEP(vp)) {
softdep_setup_allocindir_meta(nbp, ip, bp,
indirs[i - 1].in_off, nb);
bdwrite(nbp);
} else {
/*
* Write synchronously so that indirect blocks
* never point at garbage.
*/
if ((error = bwrite(nbp)) != 0) {
brelse(bp);
goto fail;
}
}
bap[indirs[i - 1].in_off] = nb;
if (allocib == NULL && unwindidx < 0)
unwindidx = i - 1;
/*
* If required, write synchronously, otherwise use
* delayed write.
*/
if (flags & B_SYNC) {
bwrite(bp);
} else {
if (bp->b_bufsize == fs->fs_bsize)
bp->b_flags |= B_CLUSTEROK;
bdwrite(bp);
}
}
/*
* Get the data block, allocating if necessary. We have already
* called getblk() on the data block buffer, dbp. If we have to
* allocate it and B_CLRBUF has been set the inference is an intention
* to zero out the related disk blocks, so we do not have to issue
* a read. Instead we simply call vfs_bio_clrbuf(). If B_CLRBUF is
* not set the caller intends to overwrite the entire contents of the
* buffer and we don't waste time trying to clean up the contents.
*
* bp references the current indirect block. When allocating,
* the block must be updated.
*/
if (nb == 0) {
pref = ffs_blkpref(ip, lbn, indirs[i].in_off, &bap[0]);
error = ffs_alloc(ip,
lbn, pref, (int)fs->fs_bsize, cred, &newb);
if (error) {
brelse(bp);
goto fail;
}
nb = newb;
*allocblk++ = nb;
*lbns_remfree++ = lbn;
dbp->b_bio2.bio_offset = fsbtodoff(fs, nb);
if (flags & B_CLRBUF)
vfs_bio_clrbuf(dbp);
if (DOINGSOFTDEP(vp))
softdep_setup_allocindir_page(ip, lbn, bp,
indirs[i].in_off, nb, 0, dbp);
bap[indirs[i].in_off] = nb;
/*
* If required, write synchronously, otherwise use
* delayed write.
*/
if (flags & B_SYNC) {
bwrite(bp);
} else {
if (bp->b_bufsize == fs->fs_bsize)
bp->b_flags |= B_CLUSTEROK;
bdwrite(bp);
}
*ap->a_bpp = dbp;
return (0);
}
brelse(bp);
/*
* At this point all related indirect blocks have been allocated
* if necessary and released. bp is no longer valid. dbp holds
* our getblk()'d data block.
*
* XXX we previously performed a cluster_read operation here.
*/
if (flags & B_CLRBUF) {
/*
* If B_CLRBUF is set we must validate the invalid portions
* of the buffer. This typically requires a read-before-
* write. The strategy call will fill in bio_offset in that
* case.
*
* If we hit this case we do a cluster read if possible
* since nearby data blocks are likely to be accessed soon
* too.
*/
if ((dbp->b_flags & B_CACHE) == 0) {
bqrelse(dbp);
seqcount = (flags & B_SEQMASK) >> B_SEQSHIFT;
if (seqcount &&
(vp->v_mount->mnt_flag & MNT_NOCLUSTERR) == 0) {
error = cluster_read(vp, (off_t)ip->i_size,
lblktodoff(fs, lbn),
(int)fs->fs_bsize,
fs->fs_bsize,
seqcount * BKVASIZE,
&dbp);
} else {
error = bread(vp, lblktodoff(fs, lbn),
(int)fs->fs_bsize, &dbp);
}
if (error)
goto fail;
} else {
/*
* Read in a superblock finding an alternate if necessary.
* Return 1 if successful, 0 if unsuccessful, -1 if filesystem
* is already clean (preen mode only).
*/
int
setup(char *dev)
{
long size, asked, i, j;
long skipclean, bmapsize;
off_t sizepb;
struct stat statb;
havesb = 0;
fswritefd = -1;
skipclean = fflag ? 0 : preen;
if (stat(dev, &statb) < 0) {
printf("Can't stat %s: %s\n", dev, strerror(errno));
return (0);
}
if ((statb.st_mode & S_IFMT) != S_IFCHR &&
(statb.st_mode & S_IFMT) != S_IFBLK) {
pfatal("%s is not a disk device", dev);
if (reply("CONTINUE") == 0)
return (0);
}
if ((fsreadfd = open(dev, O_RDONLY)) < 0) {
printf("Can't open %s: %s\n", dev, strerror(errno));
return (0);
}
if (preen == 0)
printf("** %s", dev);
if (nflag || (fswritefd = open(dev, O_WRONLY)) < 0) {
fswritefd = -1;
if (preen)
pfatal("NO WRITE ACCESS");
printf(" (NO WRITE)");
}
if (preen == 0)
printf("\n");
fsmodified = 0;
lfdir = 0;
initbarea(&sblk);
initbarea(&asblk);
sblk.b_un.b_buf = malloc(SBSIZE);
asblk.b_un.b_buf = malloc(SBSIZE);
if (sblk.b_un.b_buf == NULL || asblk.b_un.b_buf == NULL)
errx(EEXIT, "cannot allocate space for superblock");
/*
* Figure out the device block size and the sector size. The
* block size is updated by readsb() later on.
*/
{
struct partinfo pinfo;
if (ioctl(fsreadfd, DIOCGPART, &pinfo) == 0) {
dev_bsize = secsize = pinfo.media_blksize;
} else {
dev_bsize = secsize = DEV_BSIZE;
}
}
/*
* Read in the superblock, looking for alternates if necessary
*/
if (readsb(1) == 0) {
skipclean = 0;
if (bflag || preen)
return(0);
if (reply("LOOK FOR ALTERNATE SUPERBLOCKS") == 0)
return (0);
bflag = 32;
if (readsb(0) == 0) {
printf(
"YOU MUST USE THE -b OPTION TO FSCK TO SPECIFY\n"
"THE LOCATION OF AN ALTERNATE SUPER-BLOCK TO\n"
"SUPPLY NEEDED INFORMATION; SEE fsck(8).");
bflag = 0;
return(0);
}
pwarn("USING ALTERNATE SUPERBLOCK AT %d\n", bflag);
bflag = 0;
}
if (skipclean && sblock.fs_clean) {
pwarn("FILESYSTEM CLEAN; SKIPPING CHECKS\n");
return (-1);
}
maxfsblock = sblock.fs_size;
maxino = sblock.fs_ncg * sblock.fs_ipg;
/*
* Check and potentially fix certain fields in the super block.
*/
if (sblock.fs_optim != FS_OPTTIME && sblock.fs_optim != FS_OPTSPACE) {
pfatal("UNDEFINED OPTIMIZATION IN SUPERBLOCK");
if (reply("SET TO DEFAULT") == 1) {
sblock.fs_optim = FS_OPTTIME;
sbdirty();
}
}
if ((sblock.fs_minfree < 0 || sblock.fs_minfree > 99)) {
pfatal("IMPOSSIBLE MINFREE=%d IN SUPERBLOCK",
sblock.fs_minfree);
if (reply("SET TO DEFAULT") == 1) {
sblock.fs_minfree = 10;
sbdirty();
}
}
if (sblock.fs_interleave < 1 ||
sblock.fs_interleave > sblock.fs_nsect) {
pwarn("IMPOSSIBLE INTERLEAVE=%d IN SUPERBLOCK",
sblock.fs_interleave);
sblock.fs_interleave = 1;
if (preen)
printf(" (FIXED)\n");
if (preen || reply("SET TO DEFAULT") == 1) {
sbdirty();
dirty(&asblk);
}
}
if (sblock.fs_npsect < sblock.fs_nsect ||
sblock.fs_npsect > sblock.fs_nsect*2) {
pwarn("IMPOSSIBLE NPSECT=%d IN SUPERBLOCK",
sblock.fs_npsect);
sblock.fs_npsect = sblock.fs_nsect;
if (preen)
printf(" (FIXED)\n");
if (preen || reply("SET TO DEFAULT") == 1) {
sbdirty();
dirty(&asblk);
}
}
if (sblock.fs_inodefmt >= FS_44INODEFMT) {
newinofmt = 1;
} else {
sblock.fs_qbmask = ~sblock.fs_bmask;
sblock.fs_qfmask = ~sblock.fs_fmask;
/* This should match the kernel limit in ffs_oldfscompat(). */
sblock.fs_maxfilesize = (u_int64_t)1 << 39;
newinofmt = 0;
}
/*
* Convert to new inode format.
*/
if (cvtlevel >= 2 && sblock.fs_inodefmt < FS_44INODEFMT) {
if (preen)
pwarn("CONVERTING TO NEW INODE FORMAT\n");
else if (!reply("CONVERT TO NEW INODE FORMAT"))
return(0);
doinglevel2++;
sblock.fs_inodefmt = FS_44INODEFMT;
sizepb = sblock.fs_bsize;
sblock.fs_maxfilesize = sblock.fs_bsize * NDADDR - 1;
for (i = 0; i < NIADDR; i++) {
sizepb *= NINDIR(&sblock);
sblock.fs_maxfilesize += sizepb;
}
sblock.fs_maxsymlinklen = MAXSYMLINKLEN;
sblock.fs_qbmask = ~sblock.fs_bmask;
sblock.fs_qfmask = ~sblock.fs_fmask;
sbdirty();
dirty(&asblk);
}
/*
* Convert to new cylinder group format.
*/
if (cvtlevel >= 1 && sblock.fs_postblformat == FS_42POSTBLFMT) {
if (preen)
pwarn("CONVERTING TO NEW CYLINDER GROUP FORMAT\n");
else if (!reply("CONVERT TO NEW CYLINDER GROUP FORMAT"))
return(0);
doinglevel1++;
sblock.fs_postblformat = FS_DYNAMICPOSTBLFMT;
sblock.fs_nrpos = 8;
sblock.fs_postbloff =
(char *)(&sblock.fs_opostbl[0][0]) -
(char *)(&sblock.fs_firstfield);
sblock.fs_rotbloff = &sblock.fs_space[0] -
(u_char *)(&sblock.fs_firstfield);
sblock.fs_cgsize =
fragroundup(&sblock, CGSIZE(&sblock));
sbdirty();
dirty(&asblk);
}
if (asblk.b_dirty && !bflag) {
memmove(&altsblock, &sblock, (size_t)sblock.fs_sbsize);
flush(fswritefd, &asblk);
}
/*
* read in the summary info.
*/
asked = 0;
sblock.fs_csp = calloc(1, sblock.fs_cssize);
for (i = 0, j = 0; i < sblock.fs_cssize; i += sblock.fs_bsize, j++) {
size = sblock.fs_cssize - i < sblock.fs_bsize ?
sblock.fs_cssize - i : sblock.fs_bsize;
if (bread(fsreadfd, (char *)sblock.fs_csp + i,
fsbtodb(&sblock, sblock.fs_csaddr + j * sblock.fs_frag),
size) != 0 && !asked) {
pfatal("BAD SUMMARY INFORMATION");
if (reply("CONTINUE") == 0) {
ckfini(0);
exit(EEXIT);
}
asked++;
}
}
/*
* allocate and initialize the necessary maps
*/
bmapsize = roundup(howmany(maxfsblock, NBBY), sizeof(short));
blockmap = calloc((unsigned)bmapsize, sizeof (char));
if (blockmap == NULL) {
printf("cannot alloc %u bytes for blockmap\n",
(unsigned)bmapsize);
goto badsb;
}
inostathead = calloc((unsigned)(sblock.fs_ncg),
sizeof(struct inostatlist));
if (inostathead == NULL) {
printf("cannot alloc %u bytes for inostathead\n",
(unsigned)(sizeof(struct inostatlist) * (sblock.fs_ncg)));
goto badsb;
}
numdirs = sblock.fs_cstotal.cs_ndir;
/*
* Calculate the directory hash table size. Do not allocate
* a ridiculous amount of memory if we have a lot of directories.
*/
for (dirhash = 16; dirhash < numdirs; dirhash <<= 1)
;
if (dirhash > 1024*1024)
dirhash /= 8;
dirhashmask = dirhash - 1;
if (numdirs == 0) {
printf("numdirs is zero, try using an alternate superblock\n");
goto badsb;
}
inplast = 0;
listmax = numdirs + 10;
inpsort = calloc((unsigned)listmax, sizeof(struct inoinfo *));
inphead = calloc((unsigned)dirhash, sizeof(struct inoinfo *));
if (inpsort == NULL || inphead == NULL) {
printf("cannot allocate base structures for %ld directories\n",
numdirs);
goto badsb;
}
bufinit();
if (sblock.fs_flags & FS_DOSOFTDEP)
usedsoftdep = 1;
else
usedsoftdep = 0;
return (1);
badsb:
ckfini(0);
return (0);
}
void
mkfs(struct partition *pp, char *fsys, int fi, int fo, mode_t mfsmode,
uid_t mfsuid, gid_t mfsgid)
{
time_t utime;
quad_t sizepb;
int i, j, width, origdensity, fragsperinode, minfpg, optimalfpg;
int lastminfpg, mincylgrps;
long cylno, csfrags;
char tmpbuf[100]; /* XXX this will break in about 2,500 years */
if ((fsun = calloc(1, sizeof (union fs_u))) == NULL ||
(cgun = calloc(1, sizeof (union cg_u))) == NULL)
err(1, "calloc");
#ifndef STANDALONE
time(&utime);
#endif
if (mfs) {
quad_t sz = (quad_t)fssize * sectorsize;
if (sz > SIZE_T_MAX) {
errno = ENOMEM;
err(12, "mmap");
}
membase = mmap(NULL, sz, PROT_READ|PROT_WRITE,
MAP_ANON|MAP_PRIVATE, -1, (off_t)0);
if (membase == MAP_FAILED)
err(12, "mmap");
madvise(membase, sz, MADV_RANDOM);
}
fsi = fi;
fso = fo;
/*
* Validate the given file system size.
* Verify that its last block can actually be accessed.
*/
if (Oflag <= 1 && fssize > INT_MAX)
errx(13, "preposterous size %lld, max is %d", fssize, INT_MAX);
if (Oflag == 2 && fssize > MAXDISKSIZE)
errx(13, "preposterous size %lld, max is %lld", fssize,
MAXDISKSIZE);
wtfs(fssize - 1, sectorsize, (char *)&sblock);
sblock.fs_postblformat = FS_DYNAMICPOSTBLFMT;
sblock.fs_avgfilesize = avgfilesize;
sblock.fs_avgfpdir = avgfilesperdir;
/*
* Collect and verify the block and fragment sizes.
*/
if (!POWEROF2(bsize)) {
errx(16, "block size must be a power of 2, not %d", bsize);
}
if (!POWEROF2(fsize)) {
errx(17, "fragment size must be a power of 2, not %d",
fsize);
}
if (fsize < sectorsize) {
errx(18, "fragment size %d is too small, minimum is %d",
fsize, sectorsize);
}
if (bsize < MINBSIZE) {
errx(19, "block size %d is too small, minimum is %d",
bsize, MINBSIZE);
}
if (bsize > MAXBSIZE) {
errx(19, "block size %d is too large, maximum is %d",
bsize, MAXBSIZE);
}
if (bsize < fsize) {
errx(20, "block size (%d) cannot be smaller than fragment size (%d)",
bsize, fsize);
}
sblock.fs_bsize = bsize;
sblock.fs_fsize = fsize;
/*
* Calculate the superblock bitmasks and shifts.
*/
sblock.fs_bmask = ~(sblock.fs_bsize - 1);
sblock.fs_fmask = ~(sblock.fs_fsize - 1);
sblock.fs_qbmask = ~sblock.fs_bmask;
sblock.fs_qfmask = ~sblock.fs_fmask;
sblock.fs_bshift = ilog2(sblock.fs_bsize);
sblock.fs_fshift = ilog2(sblock.fs_fsize);
sblock.fs_frag = numfrags(&sblock, sblock.fs_bsize);
if (sblock.fs_frag > MAXFRAG) {
errx(21, "fragment size %d is too small, minimum with block "
"size %d is %d", sblock.fs_fsize, sblock.fs_bsize,
sblock.fs_bsize / MAXFRAG);
}
sblock.fs_fragshift = ilog2(sblock.fs_frag);
sblock.fs_fsbtodb = ilog2(sblock.fs_fsize / sectorsize);
sblock.fs_size = dbtofsb(&sblock, fssize);
sblock.fs_nspf = sblock.fs_fsize / sectorsize;
sblock.fs_maxcontig = 1;
sblock.fs_nrpos = 1;
sblock.fs_cpg = 1;
/*
* Before the file system is fully initialized, mark it as invalid.
*/
sblock.fs_magic = FS_BAD_MAGIC;
/*
* Set the remaining superblock fields. Note that for FFS1, media
* geometry fields are set to fake values. This is for compatibility
* with really ancient kernels that might still inspect these values.
*/
if (Oflag <= 1) {
sblock.fs_sblockloc = SBLOCK_UFS1;
sblock.fs_nindir = sblock.fs_bsize / sizeof(int32_t);
sblock.fs_inopb = sblock.fs_bsize / sizeof(struct ufs1_dinode);
if (Oflag == 0) {
sblock.fs_maxsymlinklen = 0;
sblock.fs_inodefmt = FS_42INODEFMT;
} else {
sblock.fs_maxsymlinklen = MAXSYMLINKLEN_UFS1;
sblock.fs_inodefmt = FS_44INODEFMT;
}
sblock.fs_cgoffset = 0;
sblock.fs_cgmask = 0xffffffff;
sblock.fs_ffs1_size = sblock.fs_size;
sblock.fs_rotdelay = 0;
sblock.fs_rps = 60;
sblock.fs_interleave = 1;
sblock.fs_trackskew = 0;
sblock.fs_cpc = 0;
} else {
sblock.fs_inodefmt = FS_44INODEFMT;
sblock.fs_sblockloc = SBLOCK_UFS2;
sblock.fs_nindir = sblock.fs_bsize / sizeof(int64_t);
sblock.fs_inopb = sblock.fs_bsize / sizeof(struct ufs2_dinode);
sblock.fs_maxsymlinklen = MAXSYMLINKLEN_UFS2;
}
sblock.fs_sblkno =
roundup(howmany(sblock.fs_sblockloc + SBLOCKSIZE, sblock.fs_fsize),
sblock.fs_frag);
sblock.fs_cblkno = (int32_t)(sblock.fs_sblkno +
roundup(howmany(SBSIZE, sblock.fs_fsize), sblock.fs_frag));
sblock.fs_iblkno = sblock.fs_cblkno + sblock.fs_frag;
sblock.fs_maxfilesize = sblock.fs_bsize * NDADDR - 1;
for (sizepb = sblock.fs_bsize, i = 0; i < NIADDR; i++) {
sizepb *= NINDIR(&sblock);
sblock.fs_maxfilesize += sizepb;
}
#ifdef notyet
/*
* It is impossible to create a snapshot in case fs_maxfilesize is
* smaller than fssize.
*/
if (sblock.fs_maxfilesize < (u_quad_t)fssize)
warnx("WARNING: You will be unable to create snapshots on this "
"file system. Correct by using a larger blocksize.");
#endif
/*
* Calculate the number of blocks to put into each cylinder group. The
* first goal is to have at least enough data blocks in each cylinder
* group to meet the density requirement. Once this goal is achieved
* we try to expand to have at least mincylgrps cylinder groups. Once
* this goal is achieved, we pack as many blocks into each cylinder
* group map as will fit.
*
* We start by calculating the smallest number of blocks that we can
* put into each cylinder group. If this is too big, we reduce the
* density until it fits.
*/
origdensity = density;
for (;;) {
fragsperinode = MAX(numfrags(&sblock, density), 1);
minfpg = fragsperinode * INOPB(&sblock);
if (minfpg > sblock.fs_size)
minfpg = sblock.fs_size;
sblock.fs_ipg = INOPB(&sblock);
sblock.fs_fpg = roundup(sblock.fs_iblkno +
sblock.fs_ipg / INOPF(&sblock), sblock.fs_frag);
if (sblock.fs_fpg < minfpg)
sblock.fs_fpg = minfpg;
sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
INOPB(&sblock));
sblock.fs_fpg = roundup(sblock.fs_iblkno +
sblock.fs_ipg / INOPF(&sblock), sblock.fs_frag);
if (sblock.fs_fpg < minfpg)
sblock.fs_fpg = minfpg;
sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
INOPB(&sblock));
if (CGSIZE(&sblock) < (unsigned long)sblock.fs_bsize)
break;
density -= sblock.fs_fsize;
}
if (density != origdensity)
warnx("density reduced from %d to %d bytes per inode",
origdensity, density);
/*
* Use a lower value for mincylgrps if the user specified a large
* number of blocks per cylinder group. This is needed for, e.g. the
* install media which needs to pack 2 files very tightly.
*/
mincylgrps = MINCYLGRPS;
if (maxfrgspercg != INT_MAX) {
i = sblock.fs_size / maxfrgspercg;
if (i < MINCYLGRPS)
mincylgrps = i <= 0 ? 1 : i;
}
/*
* Start packing more blocks into the cylinder group until it cannot
* grow any larger, the number of cylinder groups drops below
* mincylgrps, or we reach the requested size.
*/
for (;;) {
sblock.fs_fpg += sblock.fs_frag;
sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
INOPB(&sblock));
if (sblock.fs_fpg > maxfrgspercg ||
sblock.fs_size / sblock.fs_fpg < mincylgrps ||
CGSIZE(&sblock) > (unsigned long)sblock.fs_bsize)
break;
}
sblock.fs_fpg -= sblock.fs_frag;
sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
INOPB(&sblock));
if (sblock.fs_fpg > maxfrgspercg)
warnx("can't honour -c: minimum is %d", sblock.fs_fpg);
/*
* Check to be sure that the last cylinder group has enough blocks to
* be viable. If it is too small, reduce the number of blocks per
* cylinder group which will have the effect of moving more blocks into
* the last cylinder group.
*/
optimalfpg = sblock.fs_fpg;
for (;;) {
sblock.fs_ncg = howmany(sblock.fs_size, sblock.fs_fpg);
lastminfpg = roundup(sblock.fs_iblkno +
sblock.fs_ipg / INOPF(&sblock), sblock.fs_frag);
if (sblock.fs_size < lastminfpg)
errx(28, "file system size %jd < minimum size of %d",
(intmax_t)sblock.fs_size, lastminfpg);
if (sblock.fs_size % sblock.fs_fpg >= lastminfpg ||
sblock.fs_size % sblock.fs_fpg == 0)
break;
sblock.fs_fpg -= sblock.fs_frag;
sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
INOPB(&sblock));
}
if (optimalfpg != sblock.fs_fpg)
warnx("reduced number of fragments per cylinder group from %d"
" to %d to enlarge last cylinder group", optimalfpg,
sblock.fs_fpg);
/*
* Back to filling superblock fields.
*/
if (Oflag <= 1) {
sblock.fs_spc = sblock.fs_fpg * sblock.fs_nspf;
sblock.fs_nsect = sblock.fs_spc;
sblock.fs_npsect = sblock.fs_spc;
sblock.fs_ncyl = sblock.fs_ncg;
}
sblock.fs_cgsize = fragroundup(&sblock, CGSIZE(&sblock));
sblock.fs_dblkno = sblock.fs_iblkno + sblock.fs_ipg / INOPF(&sblock);
sblock.fs_csaddr = cgdmin(&sblock, 0);
sblock.fs_cssize =
fragroundup(&sblock, sblock.fs_ncg * sizeof(struct csum));
fscs = (struct csum *)calloc(1, sblock.fs_cssize);
if (fscs == NULL)
errx(31, "calloc failed");
sblock.fs_sbsize = fragroundup(&sblock, sizeof(struct fs));
if (sblock.fs_sbsize > SBLOCKSIZE)
sblock.fs_sbsize = SBLOCKSIZE;
sblock.fs_minfree = minfree;
sblock.fs_maxbpg = maxbpg;
sblock.fs_optim = opt;
sblock.fs_cgrotor = 0;
sblock.fs_pendingblocks = 0;
sblock.fs_pendinginodes = 0;
sblock.fs_fmod = 0;
sblock.fs_ronly = 0;
sblock.fs_state = 0;
sblock.fs_clean = 1;
sblock.fs_id[0] = (u_int32_t)utime;
sblock.fs_id[1] = (u_int32_t)arc4random();
sblock.fs_fsmnt[0] = '\0';
csfrags = howmany(sblock.fs_cssize, sblock.fs_fsize);
sblock.fs_dsize = sblock.fs_size - sblock.fs_sblkno -
sblock.fs_ncg * (sblock.fs_dblkno - sblock.fs_sblkno);
sblock.fs_cstotal.cs_nbfree = fragstoblks(&sblock, sblock.fs_dsize) -
howmany(csfrags, sblock.fs_frag);
sblock.fs_cstotal.cs_nffree = fragnum(&sblock, sblock.fs_size) +
(fragnum(&sblock, csfrags) > 0 ?
sblock.fs_frag - fragnum(&sblock, csfrags) : 0);
sblock.fs_cstotal.cs_nifree = sblock.fs_ncg * sblock.fs_ipg - ROOTINO;
sblock.fs_cstotal.cs_ndir = 0;
sblock.fs_dsize -= csfrags;
sblock.fs_time = utime;
if (Oflag <= 1) {
sblock.fs_ffs1_time = sblock.fs_time;
sblock.fs_ffs1_dsize = sblock.fs_dsize;
sblock.fs_ffs1_csaddr = sblock.fs_csaddr;
sblock.fs_ffs1_cstotal.cs_ndir = sblock.fs_cstotal.cs_ndir;
sblock.fs_ffs1_cstotal.cs_nbfree = sblock.fs_cstotal.cs_nbfree;
sblock.fs_ffs1_cstotal.cs_nifree = sblock.fs_cstotal.cs_nifree;
sblock.fs_ffs1_cstotal.cs_nffree = sblock.fs_cstotal.cs_nffree;
}
/*
* Dump out summary information about file system.
*/
if (!mfs) {
#define B2MBFACTOR (1 / (1024.0 * 1024.0))
printf("%s: %.1fMB in %jd sectors of %d bytes\n", fsys,
(float)sblock.fs_size * sblock.fs_fsize * B2MBFACTOR,
(intmax_t)fsbtodb(&sblock, sblock.fs_size), sectorsize);
printf("%d cylinder groups of %.2fMB, %d blocks, %d"
" inodes each\n", sblock.fs_ncg,
(float)sblock.fs_fpg * sblock.fs_fsize * B2MBFACTOR,
sblock.fs_fpg / sblock.fs_frag, sblock.fs_ipg);
#undef B2MBFACTOR
}
/*
* Wipe out old FFS1 superblock if necessary.
*/
if (Oflag >= 2) {
union fs_u *fsun1;
struct fs *fs1;
fsun1 = calloc(1, sizeof(union fs_u));
if (fsun1 == NULL)
err(39, "calloc");
fs1 = &fsun1->fs;
rdfs(SBLOCK_UFS1 / sectorsize, SBSIZE, (char *)fs1);
if (fs1->fs_magic == FS_UFS1_MAGIC) {
fs1->fs_magic = FS_BAD_MAGIC;
wtfs(SBLOCK_UFS1 / sectorsize, SBSIZE, (char *)fs1);
}
free(fsun1);
}
wtfs((int)sblock.fs_sblockloc / sectorsize, SBSIZE, (char *)&sblock);
sblock.fs_magic = (Oflag <= 1) ? FS_UFS1_MAGIC : FS_UFS2_MAGIC;
/*
* Now build the cylinders group blocks and
* then print out indices of cylinder groups.
*/
if (!quiet)
printf("super-block backups (for fsck -b #) at:\n");
#ifndef STANDALONE
else if (!mfs && isatty(STDIN_FILENO)) {
signal(SIGINFO, siginfo);
cur_fsys = fsys;
}
#endif
i = 0;
width = charsperline();
/*
* Allocate space for superblock, cylinder group map, and two sets of
* inode blocks.
*/
if (sblock.fs_bsize < SBLOCKSIZE)
iobufsize = SBLOCKSIZE + 3 * sblock.fs_bsize;
else
iobufsize = 4 * sblock.fs_bsize;
if ((iobuf = malloc(iobufsize)) == 0)
errx(38, "cannot allocate I/O buffer");
bzero(iobuf, iobufsize);
/*
* Make a copy of the superblock into the buffer that we will be
* writing out in each cylinder group.
*/
bcopy((char *)&sblock, iobuf, SBLOCKSIZE);
for (cylno = 0; cylno < sblock.fs_ncg; cylno++) {
cur_cylno = (sig_atomic_t)cylno;
initcg(cylno, utime);
if (quiet)
continue;
j = snprintf(tmpbuf, sizeof tmpbuf, " %lld,",
fsbtodb(&sblock, cgsblock(&sblock, cylno)));
if (j >= sizeof tmpbuf)
j = sizeof tmpbuf - 1;
if (j == -1 || i+j >= width) {
printf("\n");
i = 0;
}
i += j;
printf("%s", tmpbuf);
fflush(stdout);
}
if (!quiet)
printf("\n");
if (Nflag && !mfs)
exit(0);
/*
* Now construct the initial file system, then write out the superblock.
*/
if (Oflag <= 1) {
if (fsinit1(utime, mfsmode, mfsuid, mfsgid))
errx(32, "fsinit1 failed");
sblock.fs_ffs1_cstotal.cs_ndir = sblock.fs_cstotal.cs_ndir;
sblock.fs_ffs1_cstotal.cs_nbfree = sblock.fs_cstotal.cs_nbfree;
sblock.fs_ffs1_cstotal.cs_nifree = sblock.fs_cstotal.cs_nifree;
sblock.fs_ffs1_cstotal.cs_nffree = sblock.fs_cstotal.cs_nffree;
} else {
if (fsinit2(utime))
errx(32, "fsinit2 failed");
}
wtfs((int)sblock.fs_sblockloc / sectorsize, SBSIZE, (char *)&sblock);
for (i = 0; i < sblock.fs_cssize; i += sblock.fs_bsize)
wtfs(fsbtodb(&sblock, sblock.fs_csaddr + numfrags(&sblock, i)),
sblock.fs_cssize - i < sblock.fs_bsize ?
sblock.fs_cssize - i : sblock.fs_bsize,
((char *)fscs) + i);
/*
* Update information about this partion in pack label, to that it may
* be updated on disk.
*/
pp->p_fstype = FS_BSDFFS;
pp->p_fragblock =
DISKLABELV1_FFS_FRAGBLOCK(sblock.fs_fsize, sblock.fs_frag);
pp->p_cpg = sblock.fs_cpg;
}
void
pass5(void)
{
int c, blk, frags, basesize, sumsize, mapsize, savednrpos = 0;
int inomapsize, blkmapsize;
struct fs *fs = &sblock;
struct cg *cg = &cgrp;
ufs_daddr_t dbase, dmax;
ufs_daddr_t d;
long i, j, k;
struct csum *cs;
struct csum cstotal;
struct inodesc idesc[3];
char buf[MAXBSIZE];
struct cg *newcg = (struct cg *)buf;
struct ocg *ocg = (struct ocg *)buf;
inoinfo(WINO)->ino_state = USTATE;
memset(newcg, 0, (size_t)fs->fs_cgsize);
newcg->cg_niblk = fs->fs_ipg;
if (cvtlevel >= 3) {
if (fs->fs_maxcontig < 2 && fs->fs_contigsumsize > 0) {
if (preen)
pwarn("DELETING CLUSTERING MAPS\n");
if (preen || reply("DELETE CLUSTERING MAPS")) {
fs->fs_contigsumsize = 0;
doinglevel1 = 1;
sbdirty();
}
}
if (fs->fs_maxcontig > 1) {
char *doit = NULL;
if (fs->fs_contigsumsize < 1) {
doit = "CREAT";
} else if (fs->fs_contigsumsize < fs->fs_maxcontig &&
fs->fs_contigsumsize < FS_MAXCONTIG) {
doit = "EXPAND";
}
if (doit) {
i = fs->fs_contigsumsize;
fs->fs_contigsumsize =
MIN(fs->fs_maxcontig, FS_MAXCONTIG);
if (CGSIZE(fs) > fs->fs_bsize) {
pwarn("CANNOT %s CLUSTER MAPS\n", doit);
fs->fs_contigsumsize = i;
} else if (preen ||
reply("CREATE CLUSTER MAPS")) {
if (preen)
pwarn("%sING CLUSTER MAPS\n",
doit);
fs->fs_cgsize =
fragroundup(fs, CGSIZE(fs));
doinglevel1 = 1;
sbdirty();
}
}
}
}
switch ((int)fs->fs_postblformat) {
case FS_42POSTBLFMT:
basesize = (char *)(&ocg->cg_btot[0]) -
(char *)(&ocg->cg_firstfield);
sumsize = &ocg->cg_iused[0] - (u_int8_t *)(&ocg->cg_btot[0]);
mapsize = &ocg->cg_free[howmany(fs->fs_fpg, NBBY)] -
(u_char *)&ocg->cg_iused[0];
blkmapsize = howmany(fs->fs_fpg, NBBY);
inomapsize = &ocg->cg_free[0] - (u_char *)&ocg->cg_iused[0];
ocg->cg_magic = CG_MAGIC;
savednrpos = fs->fs_nrpos;
fs->fs_nrpos = 8;
break;
case FS_DYNAMICPOSTBLFMT:
newcg->cg_btotoff =
&newcg->cg_space[0] - (u_char *)(&newcg->cg_firstfield);
newcg->cg_boff =
newcg->cg_btotoff + fs->fs_cpg * sizeof(int32_t);
newcg->cg_iusedoff = newcg->cg_boff +
fs->fs_cpg * fs->fs_nrpos * sizeof(u_int16_t);
newcg->cg_freeoff =
newcg->cg_iusedoff + howmany(fs->fs_ipg, NBBY);
inomapsize = newcg->cg_freeoff - newcg->cg_iusedoff;
newcg->cg_nextfreeoff = newcg->cg_freeoff +
howmany(fs->fs_cpg * fs->fs_spc / NSPF(fs), NBBY);
blkmapsize = newcg->cg_nextfreeoff - newcg->cg_freeoff;
if (fs->fs_contigsumsize > 0) {
newcg->cg_clustersumoff = newcg->cg_nextfreeoff -
sizeof(u_int32_t);
newcg->cg_clustersumoff =
roundup(newcg->cg_clustersumoff, sizeof(u_int32_t));
newcg->cg_clusteroff = newcg->cg_clustersumoff +
(fs->fs_contigsumsize + 1) * sizeof(u_int32_t);
newcg->cg_nextfreeoff = newcg->cg_clusteroff +
howmany(fs->fs_cpg * fs->fs_spc / NSPB(fs), NBBY);
}
newcg->cg_magic = CG_MAGIC;
basesize = &newcg->cg_space[0] -
(u_char *)(&newcg->cg_firstfield);
sumsize = newcg->cg_iusedoff - newcg->cg_btotoff;
mapsize = newcg->cg_nextfreeoff - newcg->cg_iusedoff;
break;
default:
inomapsize = blkmapsize = sumsize = 0; /* keep lint happy */
errx(EEXIT, "UNKNOWN ROTATIONAL TABLE FORMAT %d",
fs->fs_postblformat);
}
memset(&idesc[0], 0, sizeof idesc);
for (i = 0; i < 3; i++) {
idesc[i].id_type = ADDR;
if (doinglevel2)
idesc[i].id_fix = FIX;
}
memset(&cstotal, 0, sizeof(struct csum));
j = blknum(fs, fs->fs_size + fs->fs_frag - 1);
for (i = fs->fs_size; i < j; i++)
setbmap(i);
for (c = 0; c < fs->fs_ncg; c++) {
if (got_siginfo) {
printf("%s: phase 5: cyl group %d of %d (%d%%)\n",
cdevname, c, sblock.fs_ncg,
c * 100 / sblock.fs_ncg);
got_siginfo = 0;
}
getblk(&cgblk, cgtod(fs, c), fs->fs_cgsize);
if (!cg_chkmagic(cg))
pfatal("CG %d: BAD MAGIC NUMBER\n", c);
dbase = cgbase(fs, c);
dmax = dbase + fs->fs_fpg;
if (dmax > fs->fs_size)
dmax = fs->fs_size;
newcg->cg_time = cg->cg_time;
newcg->cg_cgx = c;
if (c == fs->fs_ncg - 1)
newcg->cg_ncyl = fs->fs_ncyl % fs->fs_cpg;
else
newcg->cg_ncyl = fs->fs_cpg;
newcg->cg_ndblk = dmax - dbase;
if (fs->fs_contigsumsize > 0)
newcg->cg_nclusterblks = newcg->cg_ndblk / fs->fs_frag;
newcg->cg_cs.cs_ndir = 0;
newcg->cg_cs.cs_nffree = 0;
newcg->cg_cs.cs_nbfree = 0;
newcg->cg_cs.cs_nifree = fs->fs_ipg;
if ((cg->cg_rotor >= 0) && (cg->cg_rotor < newcg->cg_ndblk))
newcg->cg_rotor = cg->cg_rotor;
else
newcg->cg_rotor = 0;
if ((cg->cg_frotor >= 0) && (cg->cg_frotor < newcg->cg_ndblk))
newcg->cg_frotor = cg->cg_frotor;
else
newcg->cg_frotor = 0;
if ((cg->cg_irotor >= 0) && (cg->cg_irotor < newcg->cg_niblk))
newcg->cg_irotor = cg->cg_irotor;
else
newcg->cg_irotor = 0;
memset(&newcg->cg_frsum[0], 0, sizeof newcg->cg_frsum);
memset(&cg_blktot(newcg)[0], 0,
(size_t)(sumsize + mapsize));
if (fs->fs_postblformat == FS_42POSTBLFMT)
ocg->cg_magic = CG_MAGIC;
j = fs->fs_ipg * c;
for (i = 0; i < inostathead[c].il_numalloced; j++, i++) {
switch (inoinfo(j)->ino_state) {
case USTATE:
break;
case DSTATE:
case DCLEAR:
case DFOUND:
newcg->cg_cs.cs_ndir++;
/* fall through */
case FSTATE:
case FCLEAR:
newcg->cg_cs.cs_nifree--;
setbit(cg_inosused(newcg), i);
break;
default:
if (j < ROOTINO)
break;
errx(EEXIT, "BAD STATE %d FOR INODE I=%ld",
inoinfo(j)->ino_state, j);
}
}
if (c == 0)
for (i = 0; i < ROOTINO; i++) {
setbit(cg_inosused(newcg), i);
newcg->cg_cs.cs_nifree--;
}
for (i = 0, d = dbase;
d < dmax;
d += fs->fs_frag, i += fs->fs_frag) {
frags = 0;
for (j = 0; j < fs->fs_frag; j++) {
if (testbmap(d + j))
continue;
setbit(cg_blksfree(newcg), i + j);
frags++;
}
if (frags == fs->fs_frag) {
newcg->cg_cs.cs_nbfree++;
j = cbtocylno(fs, i);
cg_blktot(newcg)[j]++;
cg_blks(fs, newcg, j)[cbtorpos(fs, i)]++;
if (fs->fs_contigsumsize > 0)
setbit(cg_clustersfree(newcg),
i / fs->fs_frag);
} else if (frags > 0) {
newcg->cg_cs.cs_nffree += frags;
blk = blkmap(fs, cg_blksfree(newcg), i);
ffs_fragacct(fs, blk, newcg->cg_frsum, 1);
}
}
if (fs->fs_contigsumsize > 0) {
int32_t *sump = cg_clustersum(newcg);
u_char *mapp = cg_clustersfree(newcg);
int map = *mapp++;
int bit = 1;
int run = 0;
for (i = 0; i < newcg->cg_nclusterblks; i++) {
if ((map & bit) != 0) {
run++;
} else if (run != 0) {
if (run > fs->fs_contigsumsize)
run = fs->fs_contigsumsize;
sump[run]++;
run = 0;
}
if ((i & (NBBY - 1)) != (NBBY - 1)) {
bit <<= 1;
} else {
map = *mapp++;
bit = 1;
}
}
if (run != 0) {
if (run > fs->fs_contigsumsize)
run = fs->fs_contigsumsize;
sump[run]++;
}
}
cstotal.cs_nffree += newcg->cg_cs.cs_nffree;
cstotal.cs_nbfree += newcg->cg_cs.cs_nbfree;
cstotal.cs_nifree += newcg->cg_cs.cs_nifree;
cstotal.cs_ndir += newcg->cg_cs.cs_ndir;
cs = &fs->fs_cs(fs, c);
if (memcmp(&newcg->cg_cs, cs, sizeof *cs) != 0 &&
dofix(&idesc[0], "FREE BLK COUNT(S) WRONG IN SUPERBLK")) {
memmove(cs, &newcg->cg_cs, sizeof *cs);
sbdirty();
}
if (doinglevel1) {
memmove(cg, newcg, (size_t)fs->fs_cgsize);
cgdirty();
continue;
}
if ((memcmp(newcg, cg, basesize) != 0 ||
memcmp(&cg_blktot(newcg)[0],
&cg_blktot(cg)[0], sumsize) != 0) &&
dofix(&idesc[2], "SUMMARY INFORMATION BAD")) {
memmove(cg, newcg, (size_t)basesize);
memmove(&cg_blktot(cg)[0],
&cg_blktot(newcg)[0], (size_t)sumsize);
cgdirty();
}
if (usedsoftdep) {
for (i = 0; i < inomapsize; i++) {
j = cg_inosused(newcg)[i];
if ((cg_inosused(cg)[i] & j) == j)
continue;
for (k = 0; k < NBBY; k++) {
if ((j & (1 << k)) == 0)
continue;
if (cg_inosused(cg)[i] & (1 << k))
continue;
pwarn("ALLOCATED INODE %d MARKED FREE\n",
c * fs->fs_ipg + i * NBBY + k);
}
}
for (i = 0; i < blkmapsize; i++) {
j = cg_blksfree(cg)[i];
if ((cg_blksfree(newcg)[i] & j) == j)
continue;
for (k = 0; k < NBBY; k++) {
if ((j & (1 << k)) == 0)
continue;
if (cg_blksfree(newcg)[i] & (1 << k))
continue;
pwarn("ALLOCATED FRAG %d MARKED FREE\n",
c * fs->fs_fpg + i * NBBY + k);
}
}
}
if (memcmp(cg_inosused(newcg), cg_inosused(cg), mapsize) != 0 &&
dofix(&idesc[1], "BLK(S) MISSING IN BIT MAPS")) {
memmove(cg_inosused(cg), cg_inosused(newcg),
(size_t)mapsize);
cgdirty();
}
}
if (fs->fs_postblformat == FS_42POSTBLFMT)
fs->fs_nrpos = savednrpos;
if (memcmp(&cstotal, &fs->fs_cstotal, sizeof *cs) != 0
&& dofix(&idesc[0], "FREE BLK COUNT(S) WRONG IN SUPERBLK")) {
memmove(&fs->fs_cstotal, &cstotal, sizeof *cs);
fs->fs_ronly = 0;
fs->fs_fmod = 0;
sbdirty();
}
}
/*
* Truncate the inode oip to at most length size, freeing the
* disk blocks.
*/
int
ffs_truncate(struct inode *oip, off_t length, int flags, struct ucred *cred)
{
struct vnode *ovp;
daddr64_t lastblock;
daddr64_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR];
daddr64_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
struct fs *fs;
struct buf *bp;
int offset, size, level;
long count, nblocks, vflags, blocksreleased = 0;
int i, aflags, error, allerror, indirect = 0;
off_t osize;
extern int num_indirdep;
extern int max_indirdep;
if (length < 0)
return (EINVAL);
ovp = ITOV(oip);
if (ovp->v_type != VREG &&
ovp->v_type != VDIR &&
ovp->v_type != VLNK)
return (0);
if (DIP(oip, size) == length)
return (0);
if (ovp->v_type == VLNK &&
(DIP(oip, size) < ovp->v_mount->mnt_maxsymlinklen ||
(ovp->v_mount->mnt_maxsymlinklen == 0 &&
oip->i_din1->di_blocks == 0))) {
#ifdef DIAGNOSTIC
if (length != 0)
panic("ffs_truncate: partial truncate of symlink");
#endif
memset(SHORTLINK(oip), 0, (size_t) DIP(oip, size));
DIP_ASSIGN(oip, size, 0);
oip->i_flag |= IN_CHANGE | IN_UPDATE;
return (UFS_UPDATE(oip, MNT_WAIT));
}
if ((error = getinoquota(oip)) != 0)
return (error);
uvm_vnp_setsize(ovp, length);
oip->i_ci.ci_lasta = oip->i_ci.ci_clen
= oip->i_ci.ci_cstart = oip->i_ci.ci_lastw = 0;
if (DOINGSOFTDEP(ovp)) {
if (length > 0 || softdep_slowdown(ovp)) {
/*
* If a file is only partially truncated, then
* we have to clean up the data structures
* describing the allocation past the truncation
* point. Finding and deallocating those structures
* is a lot of work. Since partial truncation occurs
* rarely, we solve the problem by syncing the file
* so that it will have no data structures left.
*/
if ((error = VOP_FSYNC(ovp, cred, MNT_WAIT)) != 0)
return (error);
} else {
(void)ufs_quota_free_blocks(oip, DIP(oip, blocks),
NOCRED);
softdep_setup_freeblocks(oip, length);
(void) vinvalbuf(ovp, 0, cred, curproc, 0, 0);
oip->i_flag |= IN_CHANGE | IN_UPDATE;
return (UFS_UPDATE(oip, 0));
}
}
fs = oip->i_fs;
osize = DIP(oip, size);
/*
* Lengthen the size of the file. We must ensure that the
* last byte of the file is allocated. Since the smallest
* value of osize is 0, length will be at least 1.
*/
if (osize < length) {
if (length > fs->fs_maxfilesize)
return (EFBIG);
aflags = B_CLRBUF;
if (flags & IO_SYNC)
aflags |= B_SYNC;
error = UFS_BUF_ALLOC(oip, length - 1, 1,
cred, aflags, &bp);
if (error)
return (error);
if (bp->b_lblkno >= NDADDR)
indirect = 1;
DIP_ASSIGN(oip, size, length);
uvm_vnp_setsize(ovp, length);
(void) uvm_vnp_uncache(ovp);
if (aflags & B_SYNC)
bwrite(bp);
else
bawrite(bp);
oip->i_flag |= IN_CHANGE | IN_UPDATE;
error = UFS_UPDATE(oip, MNT_WAIT);
if (DOINGSOFTDEP(ovp) && num_indirdep > max_indirdep)
if (indirect) {
/*
* If the number of pending indirect block
* dependencies is sufficiently close to the
* maximum number of simultaneously mappable
* buffers force a sync on the vnode to prevent
* buffer cache exhaustion.
*/
VOP_FSYNC(ovp, curproc->p_ucred, MNT_WAIT);
}
return (error);
}
uvm_vnp_setsize(ovp, length);
/*
* Shorten the size of the file. If the file is not being
* truncated to a block boundary, the contents of the
* partial block following the end of the file must be
* zero'ed in case it ever becomes accessible again because
* of subsequent file growth. Directories however are not
* zero'ed as they should grow back initialized to empty.
*/
offset = blkoff(fs, length);
if (offset == 0) {
DIP_ASSIGN(oip, size, length);
} else {
lbn = lblkno(fs, length);
aflags = B_CLRBUF;
if (flags & IO_SYNC)
aflags |= B_SYNC;
error = UFS_BUF_ALLOC(oip, length - 1, 1,
cred, aflags, &bp);
if (error)
return (error);
/*
* When we are doing soft updates and the UFS_BALLOC
* above fills in a direct block hole with a full sized
* block that will be truncated down to a fragment below,
* we must flush out the block dependency with an FSYNC
* so that we do not get a soft updates inconsistency
* when we create the fragment below.
*/
if (DOINGSOFTDEP(ovp) && lbn < NDADDR &&
fragroundup(fs, blkoff(fs, length)) < fs->fs_bsize &&
(error = VOP_FSYNC(ovp, cred, MNT_WAIT)) != 0)
return (error);
DIP_ASSIGN(oip, size, length);
size = blksize(fs, oip, lbn);
(void) uvm_vnp_uncache(ovp);
if (ovp->v_type != VDIR)
bzero((char *)bp->b_data + offset,
(u_int)(size - offset));
bp->b_bcount = size;
if (aflags & B_SYNC)
bwrite(bp);
else
bawrite(bp);
}
/*
* Calculate index into inode's block list of
* last direct and indirect blocks (if any)
* which we want to keep. Lastblock is -1 when
* the file is truncated to 0.
*/
lastblock = lblkno(fs, length + fs->fs_bsize - 1) - 1;
lastiblock[SINGLE] = lastblock - NDADDR;
lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
nblocks = btodb(fs->fs_bsize);
/*
* Update file and block pointers on disk before we start freeing
* blocks. If we crash before free'ing blocks below, the blocks
* will be returned to the free list. lastiblock values are also
* normalized to -1 for calls to ffs_indirtrunc below.
*/
for (level = TRIPLE; level >= SINGLE; level--) {
oldblks[NDADDR + level] = DIP(oip, ib[level]);
if (lastiblock[level] < 0) {
DIP_ASSIGN(oip, ib[level], 0);
lastiblock[level] = -1;
}
}
for (i = 0; i < NDADDR; i++) {
oldblks[i] = DIP(oip, db[i]);
if (i > lastblock)
DIP_ASSIGN(oip, db[i], 0);
}
oip->i_flag |= IN_CHANGE | IN_UPDATE;
if ((error = UFS_UPDATE(oip, MNT_WAIT)) != 0)
allerror = error;
/*
* Having written the new inode to disk, save its new configuration
* and put back the old block pointers long enough to process them.
* Note that we save the new block configuration so we can check it
* when we are done.
*/
for (i = 0; i < NDADDR; i++) {
newblks[i] = DIP(oip, db[i]);
DIP_ASSIGN(oip, db[i], oldblks[i]);
}
for (i = 0; i < NIADDR; i++) {
newblks[NDADDR + i] = DIP(oip, ib[i]);
DIP_ASSIGN(oip, ib[i], oldblks[NDADDR + i]);
}
DIP_ASSIGN(oip, size, osize);
vflags = ((length > 0) ? V_SAVE : 0) | V_SAVEMETA;
allerror = vinvalbuf(ovp, vflags, cred, curproc, 0, 0);
/*
* Indirect blocks first.
*/
indir_lbn[SINGLE] = -NDADDR;
indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1;
indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
for (level = TRIPLE; level >= SINGLE; level--) {
bn = DIP(oip, ib[level]);
if (bn != 0) {
error = ffs_indirtrunc(oip, indir_lbn[level],
fsbtodb(fs, bn), lastiblock[level], level, &count);
if (error)
allerror = error;
blocksreleased += count;
if (lastiblock[level] < 0) {
DIP_ASSIGN(oip, ib[level], 0);
ffs_blkfree(oip, bn, fs->fs_bsize);
blocksreleased += nblocks;
}
}
if (lastiblock[level] >= 0)
goto done;
}
/*
* All whole direct blocks or frags.
*/
for (i = NDADDR - 1; i > lastblock; i--) {
long bsize;
bn = DIP(oip, db[i]);
if (bn == 0)
continue;
DIP_ASSIGN(oip, db[i], 0);
bsize = blksize(fs, oip, i);
ffs_blkfree(oip, bn, bsize);
blocksreleased += btodb(bsize);
}
if (lastblock < 0)
goto done;
/*
* Finally, look for a change in size of the
* last direct block; release any frags.
*/
bn = DIP(oip, db[lastblock]);
if (bn != 0) {
long oldspace, newspace;
/*
* Calculate amount of space we're giving
* back as old block size minus new block size.
*/
oldspace = blksize(fs, oip, lastblock);
DIP_ASSIGN(oip, size, length);
newspace = blksize(fs, oip, lastblock);
if (newspace == 0)
panic("ffs_truncate: newspace");
if (oldspace - newspace > 0) {
/*
* Block number of space to be free'd is
* the old block # plus the number of frags
* required for the storage we're keeping.
*/
bn += numfrags(fs, newspace);
ffs_blkfree(oip, bn, oldspace - newspace);
blocksreleased += btodb(oldspace - newspace);
}
}
done:
#ifdef DIAGNOSTIC
for (level = SINGLE; level <= TRIPLE; level++)
if (newblks[NDADDR + level] != DIP(oip, ib[level]))
panic("ffs_truncate1");
for (i = 0; i < NDADDR; i++)
if (newblks[i] != DIP(oip, db[i]))
panic("ffs_truncate2");
#endif /* DIAGNOSTIC */
/*
* Put back the real size.
*/
DIP_ASSIGN(oip, size, length);
DIP_ADD(oip, blocks, -blocksreleased);
oip->i_flag |= IN_CHANGE;
(void)ufs_quota_free_blocks(oip, blocksreleased, NOCRED);
return (allerror);
}
int
ckinode(struct ufs1_dinode *dp, struct inodesc *idesc)
{
ufs_daddr_t *ap;
int ret;
long n, ndb, offset;
struct ufs1_dinode dino;
quad_t remsize, sizepb;
mode_t mode;
char pathbuf[MAXPATHLEN + 1];
if (idesc->id_fix != IGNORE)
idesc->id_fix = DONTKNOW;
idesc->id_entryno = 0;
idesc->id_filesize = dp->di_size;
mode = dp->di_mode & IFMT;
if (mode == IFBLK || mode == IFCHR || (mode == IFLNK &&
dp->di_size < (unsigned)sblock.fs_maxsymlinklen))
return (KEEPON);
dino = *dp;
ndb = howmany(dino.di_size, sblock.fs_bsize);
for (ap = &dino.di_db[0]; ap < &dino.di_db[NDADDR]; ap++) {
if (--ndb == 0 && (offset = blkoff(&sblock, dino.di_size)) != 0)
idesc->id_numfrags =
numfrags(&sblock, fragroundup(&sblock, offset));
else
idesc->id_numfrags = sblock.fs_frag;
if (*ap == 0) {
if (idesc->id_type == DATA && ndb >= 0) {
/* An empty block in a directory XXX */
getpathname(pathbuf, idesc->id_number,
idesc->id_number);
pfatal("DIRECTORY %s: CONTAINS EMPTY BLOCKS",
pathbuf);
if (reply("ADJUST LENGTH") == 1) {
dp = ginode(idesc->id_number);
dp->di_size = (ap - &dino.di_db[0]) *
sblock.fs_bsize;
printf(
"YOU MUST RERUN FSCK AFTERWARDS\n");
rerun = 1;
inodirty();
}
}
continue;
}
idesc->id_blkno = *ap;
if (idesc->id_type == ADDR)
ret = (*idesc->id_func)(idesc);
else
ret = dirscan(idesc);
if (ret & STOP)
return (ret);
}
idesc->id_numfrags = sblock.fs_frag;
remsize = dino.di_size - sblock.fs_bsize * NDADDR;
sizepb = sblock.fs_bsize;
for (ap = &dino.di_ib[0], n = 1; n <= NIADDR; ap++, n++) {
if (*ap) {
idesc->id_blkno = *ap;
ret = iblock(idesc, n, remsize);
if (ret & STOP)
return (ret);
} else {
if (idesc->id_type == DATA && remsize > 0) {
/* An empty block in a directory XXX */
getpathname(pathbuf, idesc->id_number,
idesc->id_number);
pfatal("DIRECTORY %s: CONTAINS EMPTY BLOCKS",
pathbuf);
if (reply("ADJUST LENGTH") == 1) {
dp = ginode(idesc->id_number);
dp->di_size -= remsize;
remsize = 0;
printf(
"YOU MUST RERUN FSCK AFTERWARDS\n");
rerun = 1;
inodirty();
break;
}
}
}
sizepb *= NINDIR(&sblock);
remsize -= sizepb;
}
return (KEEPON);
}
/*
* Balloc defines the structure of filesystem storage
* by allocating the physical blocks on a device given
* the inode and the logical block number in a file.
*/
int
ext2_balloc(struct inode *ip, e2fs_lbn_t lbn, int size, struct ucred *cred,
struct buf **bpp, int flags)
{
struct m_ext2fs *fs;
struct ext2mount *ump;
struct buf *bp, *nbp;
struct vnode *vp = ITOV(ip);
struct indir indirs[NIADDR + 2];
e4fs_daddr_t nb, newb;
e2fs_daddr_t *bap, pref;
int osize, nsize, num, i, error;
*bpp = NULL;
if (lbn < 0)
return (EFBIG);
fs = ip->i_e2fs;
ump = ip->i_ump;
/*
* check if this is a sequential block allocation.
* If so, increment next_alloc fields to allow ext2_blkpref
* to make a good guess
*/
if (lbn == ip->i_next_alloc_block + 1) {
ip->i_next_alloc_block++;
ip->i_next_alloc_goal++;
}
/*
* The first NDADDR blocks are direct blocks
*/
if (lbn < NDADDR) {
nb = ip->i_db[lbn];
/* no new block is to be allocated, and no need to expand
the file */
if (nb != 0 && ip->i_size >= (lbn + 1) * fs->e2fs_bsize) {
error = bread(vp, lbn, fs->e2fs_bsize, NOCRED, &bp);
if (error) {
brelse(bp);
return (error);
}
bp->b_blkno = fsbtodb(fs, nb);
*bpp = bp;
return (0);
}
if (nb != 0) {
/*
* Consider need to reallocate a fragment.
*/
osize = fragroundup(fs, blkoff(fs, ip->i_size));
nsize = fragroundup(fs, size);
if (nsize <= osize) {
error = bread(vp, lbn, osize, NOCRED, &bp);
if (error) {
brelse(bp);
return (error);
}
bp->b_blkno = fsbtodb(fs, nb);
} else {
/* Godmar thinks: this shouldn't happen w/o fragments */
printf("nsize %d(%d) > osize %d(%d) nb %d\n",
(int)nsize, (int)size, (int)osize,
(int)ip->i_size, (int)nb);
panic(
"ext2_balloc: Something is terribly wrong");
/*
* please note there haven't been any changes from here on -
* FFS seems to work.
*/
}
} else {
if (ip->i_size < (lbn + 1) * fs->e2fs_bsize)
nsize = fragroundup(fs, size);
else
nsize = fs->e2fs_bsize;
EXT2_LOCK(ump);
error = ext2_alloc(ip, lbn,
ext2_blkpref(ip, lbn, (int)lbn, &ip->i_db[0], 0),
nsize, cred, &newb);
if (error)
return (error);
bp = getblk(vp, lbn, nsize, 0, 0, 0);
bp->b_blkno = fsbtodb(fs, newb);
if (flags & BA_CLRBUF)
vfs_bio_clrbuf(bp);
}
ip->i_db[lbn] = dbtofsb(fs, bp->b_blkno);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
*bpp = bp;
return (0);
}
/*
* Determine the number of levels of indirection.
*/
pref = 0;
if ((error = ext2_getlbns(vp, lbn, indirs, &num)) != 0)
return (error);
#ifdef INVARIANTS
if (num < 1)
panic ("ext2_balloc: ext2_getlbns returned indirect block");
#endif
/*
* Fetch the first indirect block allocating if necessary.
*/
--num;
nb = ip->i_ib[indirs[0].in_off];
if (nb == 0) {
EXT2_LOCK(ump);
pref = ext2_blkpref(ip, lbn, indirs[0].in_off +
EXT2_NDIR_BLOCKS, &ip->i_db[0], 0);
if ((error = ext2_alloc(ip, lbn, pref, fs->e2fs_bsize, cred,
&newb)))
return (error);
nb = newb;
bp = getblk(vp, indirs[1].in_lbn, fs->e2fs_bsize, 0, 0, 0);
bp->b_blkno = fsbtodb(fs, newb);
vfs_bio_clrbuf(bp);
/*
* Write synchronously so that indirect blocks
* never point at garbage.
*/
if ((error = bwrite(bp)) != 0) {
ext2_blkfree(ip, nb, fs->e2fs_bsize);
return (error);
}
ip->i_ib[indirs[0].in_off] = newb;
ip->i_flag |= IN_CHANGE | IN_UPDATE;
}
/*
* Fetch through the indirect blocks, allocating as necessary.
*/
for (i = 1;;) {
error = bread(vp,
indirs[i].in_lbn, (int)fs->e2fs_bsize, NOCRED, &bp);
if (error) {
brelse(bp);
return (error);
}
bap = (e2fs_daddr_t *)bp->b_data;
nb = bap[indirs[i].in_off];
if (i == num)
break;
i += 1;
if (nb != 0) {
bqrelse(bp);
continue;
}
EXT2_LOCK(ump);
if (pref == 0)
pref = ext2_blkpref(ip, lbn, indirs[i].in_off, bap,
bp->b_lblkno);
error = ext2_alloc(ip, lbn, pref, (int)fs->e2fs_bsize, cred, &newb);
if (error) {
brelse(bp);
return (error);
}
nb = newb;
nbp = getblk(vp, indirs[i].in_lbn, fs->e2fs_bsize, 0, 0, 0);
nbp->b_blkno = fsbtodb(fs, nb);
vfs_bio_clrbuf(nbp);
/*
* Write synchronously so that indirect blocks
* never point at garbage.
*/
if ((error = bwrite(nbp)) != 0) {
ext2_blkfree(ip, nb, fs->e2fs_bsize);
EXT2_UNLOCK(ump);
brelse(bp);
return (error);
}
bap[indirs[i - 1].in_off] = nb;
/*
* If required, write synchronously, otherwise use
* delayed write.
*/
if (flags & IO_SYNC) {
bwrite(bp);
} else {
if (bp->b_bufsize == fs->e2fs_bsize)
bp->b_flags |= B_CLUSTEROK;
bdwrite(bp);
}
}
/*
* Get the data block, allocating if necessary.
*/
if (nb == 0) {
EXT2_LOCK(ump);
pref = ext2_blkpref(ip, lbn, indirs[i].in_off, &bap[0],
bp->b_lblkno);
if ((error = ext2_alloc(ip,
lbn, pref, (int)fs->e2fs_bsize, cred, &newb)) != 0) {
brelse(bp);
return (error);
}
nb = newb;
nbp = getblk(vp, lbn, fs->e2fs_bsize, 0, 0, 0);
nbp->b_blkno = fsbtodb(fs, nb);
if (flags & BA_CLRBUF)
vfs_bio_clrbuf(nbp);
bap[indirs[i].in_off] = nb;
/*
* If required, write synchronously, otherwise use
* delayed write.
*/
if (flags & IO_SYNC) {
bwrite(bp);
} else {
if (bp->b_bufsize == fs->e2fs_bsize)
bp->b_flags |= B_CLUSTEROK;
bdwrite(bp);
}
*bpp = nbp;
return (0);
}
brelse(bp);
if (flags & BA_CLRBUF) {
int seqcount = (flags & BA_SEQMASK) >> BA_SEQSHIFT;
if (seqcount && (vp->v_mount->mnt_flag & MNT_NOCLUSTERR) == 0) {
error = cluster_read(vp, ip->i_size, lbn,
(int)fs->e2fs_bsize, NOCRED,
MAXBSIZE, seqcount, 0, &nbp);
} else {
error = bread(vp, lbn, (int)fs->e2fs_bsize, NOCRED, &nbp);
}
if (error) {
brelse(nbp);
return (error);
}
} else {
/*
* Check validity of held blocks in an inode, recursing through all blocks.
*/
int
ckinode(struct ufs1_dinode *dp, struct inodesc *idesc)
{
ufs_daddr_t *ap, lbn;
long ret, n, ndb, offset;
struct ufs1_dinode dino;
u_int64_t remsize, sizepb;
mode_t mode;
char pathbuf[MAXPATHLEN + 1];
struct uvnode *vp, *thisvp;
if (idesc->id_fix != IGNORE)
idesc->id_fix = DONTKNOW;
idesc->id_entryno = 0;
idesc->id_filesize = dp->di_size;
mode = dp->di_mode & IFMT;
if (mode == IFBLK || mode == IFCHR ||
(mode == IFLNK && (dp->di_size < fs->lfs_maxsymlinklen ||
(fs->lfs_maxsymlinklen == 0 &&
dp->di_blocks == 0))))
return (KEEPON);
dino = *dp;
ndb = howmany(dino.di_size, fs->lfs_bsize);
thisvp = vget(fs, idesc->id_number);
for (lbn = 0; lbn < NDADDR; lbn++) {
ap = dino.di_db + lbn;
if (thisvp)
idesc->id_numfrags =
numfrags(fs, VTOI(thisvp)->i_lfs_fragsize[lbn]);
else {
if (--ndb == 0 && (offset = blkoff(fs, dino.di_size)) != 0) {
idesc->id_numfrags =
numfrags(fs, fragroundup(fs, offset));
} else
idesc->id_numfrags = fs->lfs_frag;
}
if (*ap == 0) {
if (idesc->id_type == DATA && ndb >= 0) {
/* An empty block in a directory XXX */
getpathname(pathbuf, sizeof(pathbuf),
idesc->id_number, idesc->id_number);
pfatal("DIRECTORY %s INO %lld: CONTAINS EMPTY BLOCKS [1]",
pathbuf, (long long)idesc->id_number);
if (reply("ADJUST LENGTH") == 1) {
vp = vget(fs, idesc->id_number);
dp = VTOD(vp);
dp->di_size = (ap - &dino.di_db[0]) *
fs->lfs_bsize;
printf(
"YOU MUST RERUN FSCK AFTERWARDS\n");
rerun = 1;
inodirty(VTOI(vp));
} else
break;
}
continue;
}
idesc->id_blkno = *ap;
idesc->id_lblkno = ap - &dino.di_db[0];
if (idesc->id_type == ADDR) {
ret = (*idesc->id_func) (idesc);
} else
ret = dirscan(idesc);
if (ret & STOP)
return (ret);
}
idesc->id_numfrags = fs->lfs_frag;
remsize = dino.di_size - fs->lfs_bsize * NDADDR;
sizepb = fs->lfs_bsize;
for (ap = &dino.di_ib[0], n = 1; n <= NIADDR; ap++, n++) {
if (*ap) {
idesc->id_blkno = *ap;
ret = iblock(idesc, n, remsize);
if (ret & STOP)
return (ret);
} else {
if (idesc->id_type == DATA && remsize > 0) {
/* An empty block in a directory XXX */
getpathname(pathbuf, sizeof(pathbuf),
idesc->id_number, idesc->id_number);
pfatal("DIRECTORY %s INO %lld: CONTAINS EMPTY BLOCKS [2]",
pathbuf, (long long)idesc->id_number);
if (reply("ADJUST LENGTH") == 1) {
vp = vget(fs, idesc->id_number);
dp = VTOD(vp);
dp->di_size -= remsize;
remsize = 0;
printf(
"YOU MUST RERUN FSCK AFTERWARDS\n");
rerun = 1;
inodirty(VTOI(vp));
break;
} else
break;
}
}
sizepb *= NINDIR(fs);
remsize -= sizepb;
}
return (KEEPON);
}
void
mkfs(struct partition *pp, char *fsys)
{
int fragsperinode, optimalfpg, origdensity, minfpg, lastminfpg;
long i, j, csfrags;
uint cg;
time_t utime;
quad_t sizepb;
int width;
ino_t maxinum;
int minfragsperinode; /* minimum ratio of frags to inodes */
char tmpbuf[100]; /* XXX this will break in about 2,500 years */
union {
struct fs fdummy;
char cdummy[SBLOCKSIZE];
} dummy;
#define fsdummy dummy.fdummy
#define chdummy dummy.cdummy
/*
* Our blocks == sector size, and the version of UFS we are using is
* specified by Oflag.
*/
disk.d_bsize = sectorsize;
disk.d_ufs = Oflag;
if (Rflag) {
utime = 1000000000;
} else {
time(&utime);
arc4random_stir();
}
sblock.fs_old_flags = FS_FLAGS_UPDATED;
sblock.fs_flags = 0;
if (Uflag)
sblock.fs_flags |= FS_DOSOFTDEP;
if (Lflag)
strlcpy(sblock.fs_volname, volumelabel, MAXVOLLEN);
if (Jflag)
sblock.fs_flags |= FS_GJOURNAL;
if (lflag)
sblock.fs_flags |= FS_MULTILABEL;
if (tflag)
sblock.fs_flags |= FS_TRIM;
/*
* Validate the given file system size.
* Verify that its last block can actually be accessed.
* Convert to file system fragment sized units.
*/
if (fssize <= 0) {
printf("preposterous size %jd\n", (intmax_t)fssize);
exit(13);
}
wtfs(fssize - (realsectorsize / DEV_BSIZE), realsectorsize,
(char *)&sblock);
/*
* collect and verify the file system density info
*/
sblock.fs_avgfilesize = avgfilesize;
sblock.fs_avgfpdir = avgfilesperdir;
if (sblock.fs_avgfilesize <= 0)
printf("illegal expected average file size %d\n",
sblock.fs_avgfilesize), exit(14);
if (sblock.fs_avgfpdir <= 0)
printf("illegal expected number of files per directory %d\n",
sblock.fs_avgfpdir), exit(15);
restart:
/*
* collect and verify the block and fragment sizes
*/
sblock.fs_bsize = bsize;
sblock.fs_fsize = fsize;
if (!POWEROF2(sblock.fs_bsize)) {
printf("block size must be a power of 2, not %d\n",
sblock.fs_bsize);
exit(16);
}
if (!POWEROF2(sblock.fs_fsize)) {
printf("fragment size must be a power of 2, not %d\n",
sblock.fs_fsize);
exit(17);
}
if (sblock.fs_fsize < sectorsize) {
printf("increasing fragment size from %d to sector size (%d)\n",
sblock.fs_fsize, sectorsize);
sblock.fs_fsize = sectorsize;
}
if (sblock.fs_bsize > MAXBSIZE) {
printf("decreasing block size from %d to maximum (%d)\n",
sblock.fs_bsize, MAXBSIZE);
sblock.fs_bsize = MAXBSIZE;
}
if (sblock.fs_bsize < MINBSIZE) {
printf("increasing block size from %d to minimum (%d)\n",
sblock.fs_bsize, MINBSIZE);
sblock.fs_bsize = MINBSIZE;
}
if (sblock.fs_fsize > MAXBSIZE) {
printf("decreasing fragment size from %d to maximum (%d)\n",
sblock.fs_fsize, MAXBSIZE);
sblock.fs_fsize = MAXBSIZE;
}
if (sblock.fs_bsize < sblock.fs_fsize) {
printf("increasing block size from %d to fragment size (%d)\n",
sblock.fs_bsize, sblock.fs_fsize);
sblock.fs_bsize = sblock.fs_fsize;
}
if (sblock.fs_fsize * MAXFRAG < sblock.fs_bsize) {
printf(
"increasing fragment size from %d to block size / %d (%d)\n",
sblock.fs_fsize, MAXFRAG, sblock.fs_bsize / MAXFRAG);
sblock.fs_fsize = sblock.fs_bsize / MAXFRAG;
}
if (maxbsize == 0)
maxbsize = bsize;
if (maxbsize < bsize || !POWEROF2(maxbsize)) {
sblock.fs_maxbsize = sblock.fs_bsize;
printf("Extent size set to %d\n", sblock.fs_maxbsize);
} else if (sblock.fs_maxbsize > FS_MAXCONTIG * sblock.fs_bsize) {
sblock.fs_maxbsize = FS_MAXCONTIG * sblock.fs_bsize;
printf("Extent size reduced to %d\n", sblock.fs_maxbsize);
} else {
sblock.fs_maxbsize = maxbsize;
}
/*
* Maxcontig sets the default for the maximum number of blocks
* that may be allocated sequentially. With file system clustering
* it is possible to allocate contiguous blocks up to the maximum
* transfer size permitted by the controller or buffering.
*/
if (maxcontig == 0)
maxcontig = MAX(1, MAXPHYS / bsize);
sblock.fs_maxcontig = maxcontig;
if (sblock.fs_maxcontig < sblock.fs_maxbsize / sblock.fs_bsize) {
sblock.fs_maxcontig = sblock.fs_maxbsize / sblock.fs_bsize;
printf("Maxcontig raised to %d\n", sblock.fs_maxbsize);
}
if (sblock.fs_maxcontig > 1)
sblock.fs_contigsumsize = MIN(sblock.fs_maxcontig,FS_MAXCONTIG);
sblock.fs_bmask = ~(sblock.fs_bsize - 1);
sblock.fs_fmask = ~(sblock.fs_fsize - 1);
sblock.fs_qbmask = ~sblock.fs_bmask;
sblock.fs_qfmask = ~sblock.fs_fmask;
sblock.fs_bshift = ilog2(sblock.fs_bsize);
sblock.fs_fshift = ilog2(sblock.fs_fsize);
sblock.fs_frag = numfrags(&sblock, sblock.fs_bsize);
sblock.fs_fragshift = ilog2(sblock.fs_frag);
if (sblock.fs_frag > MAXFRAG) {
printf("fragment size %d is still too small (can't happen)\n",
sblock.fs_bsize / MAXFRAG);
exit(21);
}
sblock.fs_fsbtodb = ilog2(sblock.fs_fsize / sectorsize);
sblock.fs_size = fssize = dbtofsb(&sblock, fssize);
sblock.fs_providersize = dbtofsb(&sblock, mediasize / sectorsize);
/*
* Before the filesystem is finally initialized, mark it
* as incompletely initialized.
*/
sblock.fs_magic = FS_BAD_MAGIC;
if (Oflag == 1) {
sblock.fs_sblockloc = SBLOCK_UFS1;
sblock.fs_nindir = sblock.fs_bsize / sizeof(ufs1_daddr_t);
sblock.fs_inopb = sblock.fs_bsize / sizeof(struct ufs1_dinode);
sblock.fs_maxsymlinklen = ((NDADDR + NIADDR) *
sizeof(ufs1_daddr_t));
sblock.fs_old_inodefmt = FS_44INODEFMT;
sblock.fs_old_cgoffset = 0;
sblock.fs_old_cgmask = 0xffffffff;
sblock.fs_old_size = sblock.fs_size;
sblock.fs_old_rotdelay = 0;
sblock.fs_old_rps = 60;
sblock.fs_old_nspf = sblock.fs_fsize / sectorsize;
sblock.fs_old_cpg = 1;
sblock.fs_old_interleave = 1;
sblock.fs_old_trackskew = 0;
sblock.fs_old_cpc = 0;
sblock.fs_old_postblformat = 1;
sblock.fs_old_nrpos = 1;
} else {
sblock.fs_sblockloc = SBLOCK_UFS2;
sblock.fs_nindir = sblock.fs_bsize / sizeof(ufs2_daddr_t);
sblock.fs_inopb = sblock.fs_bsize / sizeof(struct ufs2_dinode);
sblock.fs_maxsymlinklen = ((NDADDR + NIADDR) *
sizeof(ufs2_daddr_t));
}
sblock.fs_sblkno =
roundup(howmany(sblock.fs_sblockloc + SBLOCKSIZE, sblock.fs_fsize),
sblock.fs_frag);
sblock.fs_cblkno = sblock.fs_sblkno +
roundup(howmany(SBLOCKSIZE, sblock.fs_fsize), sblock.fs_frag);
sblock.fs_iblkno = sblock.fs_cblkno + sblock.fs_frag;
sblock.fs_maxfilesize = sblock.fs_bsize * NDADDR - 1;
for (sizepb = sblock.fs_bsize, i = 0; i < NIADDR; i++) {
sizepb *= NINDIR(&sblock);
sblock.fs_maxfilesize += sizepb;
}
/*
* It's impossible to create a snapshot in case that fs_maxfilesize
* is smaller than the fssize.
*/
if (sblock.fs_maxfilesize < (u_quad_t)fssize) {
warnx("WARNING: You will be unable to create snapshots on this "
"file system. Correct by using a larger blocksize.");
}
/*
* Calculate the number of blocks to put into each cylinder group.
*
* This algorithm selects the number of blocks per cylinder
* group. The first goal is to have at least enough data blocks
* in each cylinder group to meet the density requirement. Once
* this goal is achieved we try to expand to have at least
* MINCYLGRPS cylinder groups. Once this goal is achieved, we
* pack as many blocks into each cylinder group map as will fit.
*
* We start by calculating the smallest number of blocks that we
* can put into each cylinder group. If this is too big, we reduce
* the density until it fits.
*/
maxinum = (((int64_t)(1)) << 32) - INOPB(&sblock);
minfragsperinode = 1 + fssize / maxinum;
if (density == 0) {
density = MAX(NFPI, minfragsperinode) * fsize;
} else if (density < minfragsperinode * fsize) {
origdensity = density;
density = minfragsperinode * fsize;
fprintf(stderr, "density increased from %d to %d\n",
origdensity, density);
}
origdensity = density;
for (;;) {
fragsperinode = MAX(numfrags(&sblock, density), 1);
if (fragsperinode < minfragsperinode) {
bsize <<= 1;
fsize <<= 1;
printf("Block size too small for a file system %s %d\n",
"of this size. Increasing blocksize to", bsize);
goto restart;
}
minfpg = fragsperinode * INOPB(&sblock);
if (minfpg > sblock.fs_size)
minfpg = sblock.fs_size;
sblock.fs_ipg = INOPB(&sblock);
sblock.fs_fpg = roundup(sblock.fs_iblkno +
sblock.fs_ipg / INOPF(&sblock), sblock.fs_frag);
if (sblock.fs_fpg < minfpg)
sblock.fs_fpg = minfpg;
sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
INOPB(&sblock));
sblock.fs_fpg = roundup(sblock.fs_iblkno +
sblock.fs_ipg / INOPF(&sblock), sblock.fs_frag);
if (sblock.fs_fpg < minfpg)
sblock.fs_fpg = minfpg;
sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
INOPB(&sblock));
if (CGSIZE(&sblock) < (unsigned long)sblock.fs_bsize)
break;
density -= sblock.fs_fsize;
}
if (density != origdensity)
printf("density reduced from %d to %d\n", origdensity, density);
/*
* Start packing more blocks into the cylinder group until
* it cannot grow any larger, the number of cylinder groups
* drops below MINCYLGRPS, or we reach the size requested.
* For UFS1 inodes per cylinder group are stored in an int16_t
* so fs_ipg is limited to 2^15 - 1.
*/
for ( ; sblock.fs_fpg < maxblkspercg; sblock.fs_fpg += sblock.fs_frag) {
sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
INOPB(&sblock));
if (Oflag > 1 || (Oflag == 1 && sblock.fs_ipg <= 0x7fff)) {
if (sblock.fs_size / sblock.fs_fpg < MINCYLGRPS)
break;
if (CGSIZE(&sblock) < (unsigned long)sblock.fs_bsize)
continue;
if (CGSIZE(&sblock) == (unsigned long)sblock.fs_bsize)
break;
}
sblock.fs_fpg -= sblock.fs_frag;
sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
INOPB(&sblock));
break;
}
/*
* Check to be sure that the last cylinder group has enough blocks
* to be viable. If it is too small, reduce the number of blocks
* per cylinder group which will have the effect of moving more
* blocks into the last cylinder group.
*/
optimalfpg = sblock.fs_fpg;
for (;;) {
sblock.fs_ncg = howmany(sblock.fs_size, sblock.fs_fpg);
lastminfpg = roundup(sblock.fs_iblkno +
sblock.fs_ipg / INOPF(&sblock), sblock.fs_frag);
if (sblock.fs_size < lastminfpg) {
printf("Filesystem size %jd < minimum size of %d\n",
(intmax_t)sblock.fs_size, lastminfpg);
exit(28);
}
if (sblock.fs_size % sblock.fs_fpg >= lastminfpg ||
sblock.fs_size % sblock.fs_fpg == 0)
break;
sblock.fs_fpg -= sblock.fs_frag;
sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
INOPB(&sblock));
}
if (optimalfpg != sblock.fs_fpg)
printf("Reduced frags per cylinder group from %d to %d %s\n",
optimalfpg, sblock.fs_fpg, "to enlarge last cyl group");
sblock.fs_cgsize = fragroundup(&sblock, CGSIZE(&sblock));
sblock.fs_dblkno = sblock.fs_iblkno + sblock.fs_ipg / INOPF(&sblock);
if (Oflag == 1) {
sblock.fs_old_spc = sblock.fs_fpg * sblock.fs_old_nspf;
sblock.fs_old_nsect = sblock.fs_old_spc;
sblock.fs_old_npsect = sblock.fs_old_spc;
sblock.fs_old_ncyl = sblock.fs_ncg;
}
/*
* fill in remaining fields of the super block
*/
sblock.fs_csaddr = cgdmin(&sblock, 0);
sblock.fs_cssize =
fragroundup(&sblock, sblock.fs_ncg * sizeof(struct csum));
fscs = (struct csum *)calloc(1, sblock.fs_cssize);
if (fscs == NULL)
errx(31, "calloc failed");
sblock.fs_sbsize = fragroundup(&sblock, sizeof(struct fs));
if (sblock.fs_sbsize > SBLOCKSIZE)
sblock.fs_sbsize = SBLOCKSIZE;
sblock.fs_minfree = minfree;
if (metaspace > 0 && metaspace < sblock.fs_fpg / 2)
sblock.fs_metaspace = blknum(&sblock, metaspace);
else if (metaspace != -1)
/* reserve half of minfree for metadata blocks */
sblock.fs_metaspace = blknum(&sblock,
(sblock.fs_fpg * minfree) / 200);
if (maxbpg == 0)
sblock.fs_maxbpg = MAXBLKPG(sblock.fs_bsize);
else
sblock.fs_maxbpg = maxbpg;
sblock.fs_optim = opt;
sblock.fs_cgrotor = 0;
sblock.fs_pendingblocks = 0;
sblock.fs_pendinginodes = 0;
sblock.fs_fmod = 0;
sblock.fs_ronly = 0;
sblock.fs_state = 0;
sblock.fs_clean = 1;
sblock.fs_id[0] = (long)utime;
sblock.fs_id[1] = newfs_random();
sblock.fs_fsmnt[0] = '\0';
csfrags = howmany(sblock.fs_cssize, sblock.fs_fsize);
sblock.fs_dsize = sblock.fs_size - sblock.fs_sblkno -
sblock.fs_ncg * (sblock.fs_dblkno - sblock.fs_sblkno);
sblock.fs_cstotal.cs_nbfree =
fragstoblks(&sblock, sblock.fs_dsize) -
howmany(csfrags, sblock.fs_frag);
sblock.fs_cstotal.cs_nffree =
fragnum(&sblock, sblock.fs_size) +
(fragnum(&sblock, csfrags) > 0 ?
sblock.fs_frag - fragnum(&sblock, csfrags) : 0);
sblock.fs_cstotal.cs_nifree = sblock.fs_ncg * sblock.fs_ipg - ROOTINO;
sblock.fs_cstotal.cs_ndir = 0;
sblock.fs_dsize -= csfrags;
sblock.fs_time = utime;
if (Oflag == 1) {
sblock.fs_old_time = utime;
sblock.fs_old_dsize = sblock.fs_dsize;
sblock.fs_old_csaddr = sblock.fs_csaddr;
sblock.fs_old_cstotal.cs_ndir = sblock.fs_cstotal.cs_ndir;
sblock.fs_old_cstotal.cs_nbfree = sblock.fs_cstotal.cs_nbfree;
sblock.fs_old_cstotal.cs_nifree = sblock.fs_cstotal.cs_nifree;
sblock.fs_old_cstotal.cs_nffree = sblock.fs_cstotal.cs_nffree;
}
/*
* Dump out summary information about file system.
*/
# define B2MBFACTOR (1 / (1024.0 * 1024.0))
printf("%s: %.1fMB (%jd sectors) block size %d, fragment size %d\n",
fsys, (float)sblock.fs_size * sblock.fs_fsize * B2MBFACTOR,
(intmax_t)fsbtodb(&sblock, sblock.fs_size), sblock.fs_bsize,
sblock.fs_fsize);
printf("\tusing %d cylinder groups of %.2fMB, %d blks, %d inodes.\n",
sblock.fs_ncg, (float)sblock.fs_fpg * sblock.fs_fsize * B2MBFACTOR,
sblock.fs_fpg / sblock.fs_frag, sblock.fs_ipg);
if (sblock.fs_flags & FS_DOSOFTDEP)
printf("\twith soft updates\n");
# undef B2MBFACTOR
if (Eflag && !Nflag) {
printf("Erasing sectors [%jd...%jd]\n",
sblock.fs_sblockloc / disk.d_bsize,
fsbtodb(&sblock, sblock.fs_size) - 1);
berase(&disk, sblock.fs_sblockloc / disk.d_bsize,
sblock.fs_size * sblock.fs_fsize - sblock.fs_sblockloc);
}
/*
* Wipe out old UFS1 superblock(s) if necessary.
*/
if (!Nflag && Oflag != 1) {
i = bread(&disk, part_ofs + SBLOCK_UFS1 / disk.d_bsize, chdummy, SBLOCKSIZE);
if (i == -1)
err(1, "can't read old UFS1 superblock: %s", disk.d_error);
if (fsdummy.fs_magic == FS_UFS1_MAGIC) {
fsdummy.fs_magic = 0;
bwrite(&disk, part_ofs + SBLOCK_UFS1 / disk.d_bsize,
chdummy, SBLOCKSIZE);
for (cg = 0; cg < fsdummy.fs_ncg; cg++) {
if (fsbtodb(&fsdummy, cgsblock(&fsdummy, cg)) > fssize)
break;
bwrite(&disk, part_ofs + fsbtodb(&fsdummy,
cgsblock(&fsdummy, cg)), chdummy, SBLOCKSIZE);
}
}
}
if (!Nflag)
do_sbwrite(&disk);
if (Xflag == 1) {
printf("** Exiting on Xflag 1\n");
exit(0);
}
if (Xflag == 2)
printf("** Leaving BAD MAGIC on Xflag 2\n");
else
sblock.fs_magic = (Oflag != 1) ? FS_UFS2_MAGIC : FS_UFS1_MAGIC;
/*
* Now build the cylinders group blocks and
* then print out indices of cylinder groups.
*/
printf("super-block backups (for fsck -b #) at:\n");
i = 0;
width = charsperline();
/*
* allocate space for superblock, cylinder group map, and
* two sets of inode blocks.
*/
if (sblock.fs_bsize < SBLOCKSIZE)
iobufsize = SBLOCKSIZE + 3 * sblock.fs_bsize;
else
iobufsize = 4 * sblock.fs_bsize;
if ((iobuf = calloc(1, iobufsize)) == 0) {
printf("Cannot allocate I/O buffer\n");
exit(38);
}
/*
* Make a copy of the superblock into the buffer that we will be
* writing out in each cylinder group.
*/
bcopy((char *)&sblock, iobuf, SBLOCKSIZE);
for (cg = 0; cg < sblock.fs_ncg; cg++) {
initcg(cg, utime);
j = snprintf(tmpbuf, sizeof(tmpbuf), " %jd%s",
(intmax_t)fsbtodb(&sblock, cgsblock(&sblock, cg)),
cg < (sblock.fs_ncg-1) ? "," : "");
if (j < 0)
tmpbuf[j = 0] = '\0';
if (i + j >= width) {
printf("\n");
i = 0;
}
i += j;
printf("%s", tmpbuf);
fflush(stdout);
}
printf("\n");
if (Nflag)
exit(0);
/*
* Now construct the initial file system,
* then write out the super-block.
*/
fsinit(utime);
if (Oflag == 1) {
sblock.fs_old_cstotal.cs_ndir = sblock.fs_cstotal.cs_ndir;
sblock.fs_old_cstotal.cs_nbfree = sblock.fs_cstotal.cs_nbfree;
sblock.fs_old_cstotal.cs_nifree = sblock.fs_cstotal.cs_nifree;
sblock.fs_old_cstotal.cs_nffree = sblock.fs_cstotal.cs_nffree;
}
if (Xflag == 3) {
printf("** Exiting on Xflag 3\n");
exit(0);
}
if (!Nflag) {
do_sbwrite(&disk);
/*
* For UFS1 filesystems with a blocksize of 64K, the first
* alternate superblock resides at the location used for
* the default UFS2 superblock. As there is a valid
* superblock at this location, the boot code will use
* it as its first choice. Thus we have to ensure that
* all of its statistcs on usage are correct.
*/
if (Oflag == 1 && sblock.fs_bsize == 65536)
wtfs(fsbtodb(&sblock, cgsblock(&sblock, 0)),
sblock.fs_bsize, (char *)&sblock);
}
for (i = 0; i < sblock.fs_cssize; i += sblock.fs_bsize)
wtfs(fsbtodb(&sblock, sblock.fs_csaddr + numfrags(&sblock, i)),
sblock.fs_cssize - i < sblock.fs_bsize ?
sblock.fs_cssize - i : sblock.fs_bsize,
((char *)fscs) + i);
/*
* Update information about this partition in pack
* label, to that it may be updated on disk.
*/
if (pp != NULL) {
pp->p_fstype = FS_BSDFFS;
pp->p_fsize = sblock.fs_fsize;
pp->p_frag = sblock.fs_frag;
pp->p_cpg = sblock.fs_fpg;
}
}
pass5()
{
int c, blk, frags, basesize, sumsize, mapsize, savednrpos;
register struct fs *fs = &sblock;
register struct cg *cg = &cgrp;
daddr_t dbase, dmax;
register daddr_t d;
register long i, j;
struct csum *cs;
time_t now;
struct csum cstotal;
struct inodesc idesc;
char buf[MAXBSIZE];
register struct cg *newcg = (struct cg *)buf;
struct ocg *ocg = (struct ocg *)buf;
bzero((char *)newcg, fs->fs_cgsize);
newcg->cg_niblk = fs->fs_ipg;
switch (fs->fs_postblformat) {
case FS_42POSTBLFMT:
basesize = (char *)(&ocg->cg_btot[0]) - (char *)(&ocg->cg_link);
sumsize = &ocg->cg_iused[0] - (char *)(&ocg->cg_btot[0]);
mapsize = &ocg->cg_free[howmany(fs->fs_fpg, NBBY)] -
(u_char *)&ocg->cg_iused[0];
ocg->cg_magic = CG_MAGIC;
savednrpos = fs->fs_nrpos;
fs->fs_nrpos = 8;
break;
case FS_DYNAMICPOSTBLFMT:
newcg->cg_btotoff =
&newcg->cg_space[0] - (u_char *)(&newcg->cg_link);
newcg->cg_boff =
newcg->cg_btotoff + fs->fs_cpg * sizeof(long);
newcg->cg_iusedoff = newcg->cg_boff +
fs->fs_cpg * fs->fs_nrpos * sizeof(short);
newcg->cg_freeoff =
newcg->cg_iusedoff + howmany(fs->fs_ipg, NBBY);
newcg->cg_nextfreeoff = newcg->cg_freeoff +
howmany(fs->fs_cpg * fs->fs_spc / NSPF(fs),
NBBY);
newcg->cg_magic = CG_MAGIC;
basesize = &newcg->cg_space[0] - (u_char *)(&newcg->cg_link);
sumsize = newcg->cg_iusedoff - newcg->cg_btotoff;
mapsize = newcg->cg_nextfreeoff - newcg->cg_iusedoff;
break;
default:
errexit("UNKNOWN ROTATIONAL TABLE FORMAT %d\n",
fs->fs_postblformat);
}
bzero((char *)&idesc, sizeof(struct inodesc));
idesc.id_type = ADDR;
bzero((char *)&cstotal, sizeof(struct csum));
(void)time(&now);
for (i = fs->fs_size; i < fragroundup(fs, fs->fs_size); i++)
setbmap(i);
for (c = 0; c < fs->fs_ncg; c++) {
getblk(&cgblk, cgtod(fs, c), fs->fs_cgsize);
if (!cg_chkmagic(cg))
pfatal("CG %d: BAD MAGIC NUMBER\n", c);
dbase = cgbase(fs, c);
dmax = dbase + fs->fs_fpg;
if (dmax > fs->fs_size)
dmax = fs->fs_size;
if (now > cg->cg_time)
newcg->cg_time = cg->cg_time;
else
newcg->cg_time = now;
newcg->cg_cgx = c;
if (c == fs->fs_ncg - 1)
newcg->cg_ncyl = fs->fs_ncyl % fs->fs_cpg;
else
newcg->cg_ncyl = fs->fs_cpg;
newcg->cg_ndblk = dmax - dbase;
newcg->cg_cs.cs_ndir = 0;
newcg->cg_cs.cs_nffree = 0;
newcg->cg_cs.cs_nbfree = 0;
newcg->cg_cs.cs_nifree = fs->fs_ipg;
if (cg->cg_rotor < newcg->cg_ndblk)
newcg->cg_rotor = cg->cg_rotor;
else
newcg->cg_rotor = 0;
if (cg->cg_frotor < newcg->cg_ndblk)
newcg->cg_frotor = cg->cg_frotor;
else
newcg->cg_frotor = 0;
if (cg->cg_irotor < newcg->cg_niblk)
newcg->cg_irotor = cg->cg_irotor;
else
newcg->cg_irotor = 0;
bzero((char *)&newcg->cg_frsum[0], sizeof newcg->cg_frsum);
bzero((char *)&cg_blktot(newcg)[0], sumsize + mapsize);
if (fs->fs_postblformat == FS_42POSTBLFMT)
ocg->cg_magic = CG_MAGIC;
j = fs->fs_ipg * c;
for (i = 0; i < fs->fs_ipg; j++, i++) {
switch (statemap[j]) {
case USTATE:
break;
case DSTATE:
case DCLEAR:
case DFOUND:
newcg->cg_cs.cs_ndir++;
/* fall through */
case FSTATE:
case FCLEAR:
newcg->cg_cs.cs_nifree--;
setbit(cg_inosused(newcg), i);
break;
default:
if (j < ROOTINO)
break;
errexit("BAD STATE %d FOR INODE I=%d",
statemap[j], j);
}
}
if (c == 0)
for (i = 0; i < ROOTINO; i++) {
setbit(cg_inosused(newcg), i);
newcg->cg_cs.cs_nifree--;
}
for (i = 0, d = dbase;
d < dmax;
d += fs->fs_frag, i += fs->fs_frag) {
frags = 0;
for (j = 0; j < fs->fs_frag; j++) {
if (getbmap(d + j))
continue;
setbit(cg_blksfree(newcg), i + j);
frags++;
}
if (frags == fs->fs_frag) {
newcg->cg_cs.cs_nbfree++;
j = cbtocylno(fs, i);
cg_blktot(newcg)[j]++;
cg_blks(fs, newcg, j)[cbtorpos(fs, i)]++;
} else if (frags > 0) {
newcg->cg_cs.cs_nffree += frags;
blk = blkmap(fs, cg_blksfree(newcg), i);
fragacct(fs, blk, newcg->cg_frsum, 1);
}
}
cstotal.cs_nffree += newcg->cg_cs.cs_nffree;
cstotal.cs_nbfree += newcg->cg_cs.cs_nbfree;
cstotal.cs_nifree += newcg->cg_cs.cs_nifree;
cstotal.cs_ndir += newcg->cg_cs.cs_ndir;
cs = &fs->fs_cs(fs, c);
if (bcmp((char *)&newcg->cg_cs, (char *)cs, sizeof *cs) != 0 &&
dofix(&idesc, "FREE BLK COUNT(S) WRONG IN SUPERBLK")) {
bcopy((char *)&newcg->cg_cs, (char *)cs, sizeof *cs);
sbdirty();
}
if (cvtflag) {
bcopy((char *)newcg, (char *)cg, fs->fs_cgsize);
cgdirty();
continue;
}
if (bcmp(cg_inosused(newcg),
cg_inosused(cg), mapsize) != 0 &&
dofix(&idesc, "BLK(S) MISSING IN BIT MAPS")) {
bcopy(cg_inosused(newcg), cg_inosused(cg), mapsize);
cgdirty();
}
if ((bcmp((char *)newcg, (char *)cg, basesize) != 0 ||
bcmp((char *)&cg_blktot(newcg)[0],
(char *)&cg_blktot(cg)[0], sumsize) != 0) &&
dofix(&idesc, "SUMMARY INFORMATION BAD")) {
bcopy((char *)newcg, (char *)cg, basesize);
bcopy((char *)&cg_blktot(newcg)[0],
(char *)&cg_blktot(cg)[0], sumsize);
cgdirty();
}
}
if (fs->fs_postblformat == FS_42POSTBLFMT)
fs->fs_nrpos = savednrpos;
if (bcmp((char *)&cstotal, (char *)&fs->fs_cstotal, sizeof *cs) != 0
&& dofix(&idesc, "FREE BLK COUNT(S) WRONG IN SUPERBLK")) {
bcopy((char *)&cstotal, (char *)&fs->fs_cstotal, sizeof *cs);
fs->fs_ronly = 0;
fs->fs_fmod = 0;
sbdirty();
}
}
struct fs *
ffs_mkfs(const char *fsys, const fsinfo_t *fsopts, time_t tstamp)
{
int fragsperinode, optimalfpg, origdensity, minfpg, lastminfpg;
int32_t cylno, i, csfrags;
long long sizepb;
void *space;
int size, blks;
int nprintcols, printcolwidth;
ffs_opt_t *ffs_opts = fsopts->fs_specific;
Oflag = ffs_opts->version;
fssize = fsopts->size / fsopts->sectorsize;
sectorsize = fsopts->sectorsize;
fsize = ffs_opts->fsize;
bsize = ffs_opts->bsize;
maxbsize = ffs_opts->maxbsize;
maxblkspercg = ffs_opts->maxblkspercg;
minfree = ffs_opts->minfree;
opt = ffs_opts->optimization;
density = ffs_opts->density;
maxcontig = ffs_opts->maxcontig;
maxbpg = ffs_opts->maxbpg;
avgfilesize = ffs_opts->avgfilesize;
avgfpdir = ffs_opts->avgfpdir;
bbsize = BBSIZE;
sbsize = SBLOCKSIZE;
strlcpy(sblock.fs_volname, ffs_opts->label, sizeof(sblock.fs_volname));
if (Oflag == 0) {
sblock.fs_old_inodefmt = FS_42INODEFMT;
sblock.fs_maxsymlinklen = 0;
sblock.fs_old_flags = 0;
} else {
sblock.fs_old_inodefmt = FS_44INODEFMT;
sblock.fs_maxsymlinklen = (Oflag == 1 ? MAXSYMLINKLEN_UFS1 :
MAXSYMLINKLEN_UFS2);
sblock.fs_old_flags = FS_FLAGS_UPDATED;
sblock.fs_flags = 0;
}
/*
* Validate the given file system size.
* Verify that its last block can actually be accessed.
* Convert to file system fragment sized units.
*/
if (fssize <= 0) {
printf("preposterous size %lld\n", (long long)fssize);
exit(13);
}
ffs_wtfs(fssize - 1, sectorsize, (char *)&sblock, fsopts);
/*
* collect and verify the filesystem density info
*/
sblock.fs_avgfilesize = avgfilesize;
sblock.fs_avgfpdir = avgfpdir;
if (sblock.fs_avgfilesize <= 0)
printf("illegal expected average file size %d\n",
sblock.fs_avgfilesize), exit(14);
if (sblock.fs_avgfpdir <= 0)
printf("illegal expected number of files per directory %d\n",
sblock.fs_avgfpdir), exit(15);
/*
* collect and verify the block and fragment sizes
*/
sblock.fs_bsize = bsize;
sblock.fs_fsize = fsize;
if (!POWEROF2(sblock.fs_bsize)) {
printf("block size must be a power of 2, not %d\n",
sblock.fs_bsize);
exit(16);
}
if (!POWEROF2(sblock.fs_fsize)) {
printf("fragment size must be a power of 2, not %d\n",
sblock.fs_fsize);
exit(17);
}
if (sblock.fs_fsize < sectorsize) {
printf("fragment size %d is too small, minimum is %d\n",
sblock.fs_fsize, sectorsize);
exit(18);
}
if (sblock.fs_bsize < MINBSIZE) {
printf("block size %d is too small, minimum is %d\n",
sblock.fs_bsize, MINBSIZE);
exit(19);
}
if (sblock.fs_bsize > FFS_MAXBSIZE) {
printf("block size %d is too large, maximum is %d\n",
sblock.fs_bsize, FFS_MAXBSIZE);
exit(19);
}
if (sblock.fs_bsize < sblock.fs_fsize) {
printf("block size (%d) cannot be smaller than fragment size (%d)\n",
sblock.fs_bsize, sblock.fs_fsize);
exit(20);
}
if (maxbsize < bsize || !POWEROF2(maxbsize)) {
sblock.fs_maxbsize = sblock.fs_bsize;
printf("Extent size set to %d\n", sblock.fs_maxbsize);
} else if (sblock.fs_maxbsize > FS_MAXCONTIG * sblock.fs_bsize) {
sblock.fs_maxbsize = FS_MAXCONTIG * sblock.fs_bsize;
printf("Extent size reduced to %d\n", sblock.fs_maxbsize);
} else {
sblock.fs_maxbsize = maxbsize;
}
sblock.fs_maxcontig = maxcontig;
if (sblock.fs_maxcontig < sblock.fs_maxbsize / sblock.fs_bsize) {
sblock.fs_maxcontig = sblock.fs_maxbsize / sblock.fs_bsize;
printf("Maxcontig raised to %d\n", sblock.fs_maxbsize);
}
if (sblock.fs_maxcontig > 1)
sblock.fs_contigsumsize = MIN(sblock.fs_maxcontig,FS_MAXCONTIG);
sblock.fs_bmask = ~(sblock.fs_bsize - 1);
sblock.fs_fmask = ~(sblock.fs_fsize - 1);
sblock.fs_qbmask = ~sblock.fs_bmask;
sblock.fs_qfmask = ~sblock.fs_fmask;
for (sblock.fs_bshift = 0, i = sblock.fs_bsize; i > 1; i >>= 1)
sblock.fs_bshift++;
for (sblock.fs_fshift = 0, i = sblock.fs_fsize; i > 1; i >>= 1)
sblock.fs_fshift++;
sblock.fs_frag = numfrags(&sblock, sblock.fs_bsize);
for (sblock.fs_fragshift = 0, i = sblock.fs_frag; i > 1; i >>= 1)
sblock.fs_fragshift++;
if (sblock.fs_frag > MAXFRAG) {
printf("fragment size %d is too small, "
"minimum with block size %d is %d\n",
sblock.fs_fsize, sblock.fs_bsize,
sblock.fs_bsize / MAXFRAG);
exit(21);
}
sblock.fs_fsbtodb = ilog2(sblock.fs_fsize / sectorsize);
sblock.fs_size = sblock.fs_providersize = fssize =
dbtofsb(&sblock, fssize);
if (Oflag <= 1) {
sblock.fs_magic = FS_UFS1_MAGIC;
sblock.fs_sblockloc = SBLOCK_UFS1;
sblock.fs_nindir = sblock.fs_bsize / sizeof(ufs1_daddr_t);
sblock.fs_inopb = sblock.fs_bsize / sizeof(struct ufs1_dinode);
sblock.fs_maxsymlinklen = ((NDADDR + NIADDR) *
sizeof (ufs1_daddr_t));
sblock.fs_old_inodefmt = FS_44INODEFMT;
sblock.fs_old_cgoffset = 0;
sblock.fs_old_cgmask = 0xffffffff;
sblock.fs_old_size = sblock.fs_size;
sblock.fs_old_rotdelay = 0;
sblock.fs_old_rps = 60;
sblock.fs_old_nspf = sblock.fs_fsize / sectorsize;
sblock.fs_old_cpg = 1;
sblock.fs_old_interleave = 1;
sblock.fs_old_trackskew = 0;
sblock.fs_old_cpc = 0;
sblock.fs_old_postblformat = 1;
sblock.fs_old_nrpos = 1;
} else {
sblock.fs_magic = FS_UFS2_MAGIC;
sblock.fs_sblockloc = SBLOCK_UFS2;
sblock.fs_nindir = sblock.fs_bsize / sizeof(ufs2_daddr_t);
sblock.fs_inopb = sblock.fs_bsize / sizeof(struct ufs2_dinode);
sblock.fs_maxsymlinklen = ((NDADDR + NIADDR) *
sizeof (ufs2_daddr_t));
}
sblock.fs_sblkno =
roundup(howmany(sblock.fs_sblockloc + SBLOCKSIZE, sblock.fs_fsize),
sblock.fs_frag);
sblock.fs_cblkno = (daddr_t)(sblock.fs_sblkno +
roundup(howmany(SBLOCKSIZE, sblock.fs_fsize), sblock.fs_frag));
sblock.fs_iblkno = sblock.fs_cblkno + sblock.fs_frag;
sblock.fs_maxfilesize = sblock.fs_bsize * NDADDR - 1;
for (sizepb = sblock.fs_bsize, i = 0; i < NIADDR; i++) {
sizepb *= NINDIR(&sblock);
sblock.fs_maxfilesize += sizepb;
}
/*
* Calculate the number of blocks to put into each cylinder group.
*
* This algorithm selects the number of blocks per cylinder
* group. The first goal is to have at least enough data blocks
* in each cylinder group to meet the density requirement. Once
* this goal is achieved we try to expand to have at least
* 1 cylinder group. Once this goal is achieved, we pack as
* many blocks into each cylinder group map as will fit.
*
* We start by calculating the smallest number of blocks that we
* can put into each cylinder group. If this is too big, we reduce
* the density until it fits.
*/
origdensity = density;
for (;;) {
fragsperinode = MAX(numfrags(&sblock, density), 1);
minfpg = fragsperinode * INOPB(&sblock);
if (minfpg > sblock.fs_size)
minfpg = sblock.fs_size;
sblock.fs_ipg = INOPB(&sblock);
sblock.fs_fpg = roundup(sblock.fs_iblkno +
sblock.fs_ipg / INOPF(&sblock), sblock.fs_frag);
if (sblock.fs_fpg < minfpg)
sblock.fs_fpg = minfpg;
sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
INOPB(&sblock));
sblock.fs_fpg = roundup(sblock.fs_iblkno +
sblock.fs_ipg / INOPF(&sblock), sblock.fs_frag);
if (sblock.fs_fpg < minfpg)
sblock.fs_fpg = minfpg;
sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
INOPB(&sblock));
if (CGSIZE(&sblock) < (unsigned long)sblock.fs_bsize)
break;
density -= sblock.fs_fsize;
}
if (density != origdensity)
printf("density reduced from %d to %d\n", origdensity, density);
if (maxblkspercg <= 0 || maxblkspercg >= fssize)
maxblkspercg = fssize - 1;
/*
* Start packing more blocks into the cylinder group until
* it cannot grow any larger, the number of cylinder groups
* drops below 1, or we reach the size requested.
*/
for ( ; sblock.fs_fpg < maxblkspercg; sblock.fs_fpg += sblock.fs_frag) {
sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
INOPB(&sblock));
if (sblock.fs_size / sblock.fs_fpg < 1)
break;
if (CGSIZE(&sblock) < (unsigned long)sblock.fs_bsize)
continue;
if (CGSIZE(&sblock) == (unsigned long)sblock.fs_bsize)
break;
sblock.fs_fpg -= sblock.fs_frag;
sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
INOPB(&sblock));
break;
}
/*
* Check to be sure that the last cylinder group has enough blocks
* to be viable. If it is too small, reduce the number of blocks
* per cylinder group which will have the effect of moving more
* blocks into the last cylinder group.
*/
optimalfpg = sblock.fs_fpg;
for (;;) {
sblock.fs_ncg = howmany(sblock.fs_size, sblock.fs_fpg);
lastminfpg = roundup(sblock.fs_iblkno +
sblock.fs_ipg / INOPF(&sblock), sblock.fs_frag);
if (sblock.fs_size < lastminfpg) {
printf("Filesystem size %lld < minimum size of %d\n",
(long long)sblock.fs_size, lastminfpg);
exit(28);
}
if (sblock.fs_size % sblock.fs_fpg >= lastminfpg ||
sblock.fs_size % sblock.fs_fpg == 0)
break;
sblock.fs_fpg -= sblock.fs_frag;
sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
INOPB(&sblock));
}
if (optimalfpg != sblock.fs_fpg)
printf("Reduced frags per cylinder group from %d to %d %s\n",
optimalfpg, sblock.fs_fpg, "to enlarge last cyl group");
sblock.fs_cgsize = fragroundup(&sblock, CGSIZE(&sblock));
sblock.fs_dblkno = sblock.fs_iblkno + sblock.fs_ipg / INOPF(&sblock);
if (Oflag <= 1) {
sblock.fs_old_spc = sblock.fs_fpg * sblock.fs_old_nspf;
sblock.fs_old_nsect = sblock.fs_old_spc;
sblock.fs_old_npsect = sblock.fs_old_spc;
sblock.fs_old_ncyl = sblock.fs_ncg;
}
/*
* fill in remaining fields of the super block
*/
sblock.fs_csaddr = cgdmin(&sblock, 0);
sblock.fs_cssize =
fragroundup(&sblock, sblock.fs_ncg * sizeof(struct csum));
/*
* Setup memory for temporary in-core cylgroup summaries.
* Cribbed from ffs_mountfs().
*/
size = sblock.fs_cssize;
blks = howmany(size, sblock.fs_fsize);
if (sblock.fs_contigsumsize > 0)
size += sblock.fs_ncg * sizeof(int32_t);
if ((space = (char *)calloc(1, size)) == NULL)
err(1, "memory allocation error for cg summaries");
sblock.fs_csp = space;
space = (char *)space + sblock.fs_cssize;
if (sblock.fs_contigsumsize > 0) {
int32_t *lp;
sblock.fs_maxcluster = lp = space;
for (i = 0; i < sblock.fs_ncg; i++)
*lp++ = sblock.fs_contigsumsize;
}
sblock.fs_sbsize = fragroundup(&sblock, sizeof(struct fs));
if (sblock.fs_sbsize > SBLOCKSIZE)
sblock.fs_sbsize = SBLOCKSIZE;
sblock.fs_minfree = minfree;
sblock.fs_maxcontig = maxcontig;
sblock.fs_maxbpg = maxbpg;
sblock.fs_optim = opt;
sblock.fs_cgrotor = 0;
sblock.fs_pendingblocks = 0;
sblock.fs_pendinginodes = 0;
sblock.fs_cstotal.cs_ndir = 0;
sblock.fs_cstotal.cs_nbfree = 0;
sblock.fs_cstotal.cs_nifree = 0;
sblock.fs_cstotal.cs_nffree = 0;
sblock.fs_fmod = 0;
sblock.fs_ronly = 0;
sblock.fs_state = 0;
sblock.fs_clean = FS_ISCLEAN;
sblock.fs_ronly = 0;
sblock.fs_id[0] = tstamp;
sblock.fs_id[1] = random();
sblock.fs_fsmnt[0] = '\0';
csfrags = howmany(sblock.fs_cssize, sblock.fs_fsize);
sblock.fs_dsize = sblock.fs_size - sblock.fs_sblkno -
sblock.fs_ncg * (sblock.fs_dblkno - sblock.fs_sblkno);
sblock.fs_cstotal.cs_nbfree =
fragstoblks(&sblock, sblock.fs_dsize) -
howmany(csfrags, sblock.fs_frag);
sblock.fs_cstotal.cs_nffree =
fragnum(&sblock, sblock.fs_size) +
(fragnum(&sblock, csfrags) > 0 ?
sblock.fs_frag - fragnum(&sblock, csfrags) : 0);
sblock.fs_cstotal.cs_nifree = sblock.fs_ncg * sblock.fs_ipg - ROOTINO;
sblock.fs_cstotal.cs_ndir = 0;
sblock.fs_dsize -= csfrags;
sblock.fs_time = tstamp;
if (Oflag <= 1) {
sblock.fs_old_time = tstamp;
sblock.fs_old_dsize = sblock.fs_dsize;
sblock.fs_old_csaddr = sblock.fs_csaddr;
sblock.fs_old_cstotal.cs_ndir = sblock.fs_cstotal.cs_ndir;
sblock.fs_old_cstotal.cs_nbfree = sblock.fs_cstotal.cs_nbfree;
sblock.fs_old_cstotal.cs_nifree = sblock.fs_cstotal.cs_nifree;
sblock.fs_old_cstotal.cs_nffree = sblock.fs_cstotal.cs_nffree;
}
/*
* Dump out summary information about file system.
*/
#define B2MBFACTOR (1 / (1024.0 * 1024.0))
printf("%s: %.1fMB (%lld sectors) block size %d, "
"fragment size %d\n",
fsys, (float)sblock.fs_size * sblock.fs_fsize * B2MBFACTOR,
(long long)fsbtodb(&sblock, sblock.fs_size),
sblock.fs_bsize, sblock.fs_fsize);
printf("\tusing %d cylinder groups of %.2fMB, %d blks, "
"%d inodes.\n",
sblock.fs_ncg,
(float)sblock.fs_fpg * sblock.fs_fsize * B2MBFACTOR,
sblock.fs_fpg / sblock.fs_frag, sblock.fs_ipg);
#undef B2MBFACTOR
/*
* Now determine how wide each column will be, and calculate how
* many columns will fit in a 76 char line. 76 is the width of the
* subwindows in sysinst.
*/
printcolwidth = count_digits(
fsbtodb(&sblock, cgsblock(&sblock, sblock.fs_ncg -1)));
nprintcols = 76 / (printcolwidth + 2);
/*
* allocate space for superblock, cylinder group map, and
* two sets of inode blocks.
*/
if (sblock.fs_bsize < SBLOCKSIZE)
iobufsize = SBLOCKSIZE + 3 * sblock.fs_bsize;
else
iobufsize = 4 * sblock.fs_bsize;
if ((iobuf = malloc(iobufsize)) == NULL) {
printf("Cannot allocate I/O buffer\n");
exit(38);
}
memset(iobuf, 0, iobufsize);
/*
* Make a copy of the superblock into the buffer that we will be
* writing out in each cylinder group.
*/
memcpy(writebuf, &sblock, sbsize);
if (fsopts->needswap)
ffs_sb_swap(&sblock, (struct fs*)writebuf);
memcpy(iobuf, writebuf, SBLOCKSIZE);
printf("super-block backups (for fsck -b #) at:");
for (cylno = 0; cylno < sblock.fs_ncg; cylno++) {
initcg(cylno, tstamp, fsopts);
if (cylno % nprintcols == 0)
printf("\n");
printf(" %*lld,", printcolwidth,
(long long)fsbtodb(&sblock, cgsblock(&sblock, cylno)));
fflush(stdout);
}
printf("\n");
/*
* Now construct the initial file system,
* then write out the super-block.
*/
sblock.fs_time = tstamp;
if (Oflag <= 1) {
sblock.fs_old_cstotal.cs_ndir = sblock.fs_cstotal.cs_ndir;
sblock.fs_old_cstotal.cs_nbfree = sblock.fs_cstotal.cs_nbfree;
sblock.fs_old_cstotal.cs_nifree = sblock.fs_cstotal.cs_nifree;
sblock.fs_old_cstotal.cs_nffree = sblock.fs_cstotal.cs_nffree;
}
if (fsopts->needswap)
sblock.fs_flags |= FS_SWAPPED;
ffs_write_superblock(&sblock, fsopts);
return (&sblock);
}
/*
* Truncate the inode ip to at most length size, freeing the
* disk blocks.
*/
int
ffs_truncate(vnode *vp, off_t length, int flags, Ucred *cred)
{
print("HARVEY TODO: %s\n", __func__);
#if 0
struct inode *ip;
ufs2_daddr_t bn, lbn, lastblock, lastiblock[UFS_NIADDR];
ufs2_daddr_t indir_lbn[UFS_NIADDR], oldblks[UFS_NDADDR + UFS_NIADDR];
ufs2_daddr_t newblks[UFS_NDADDR + UFS_NIADDR];
ufs2_daddr_t count, blocksreleased = 0, datablocks, blkno;
struct bufobj *bo;
struct fs *fs;
struct buf *bp;
struct ufsmount *ump;
int softdeptrunc, journaltrunc;
int needextclean, extblocks;
int offset, size, level, nblocks;
int i, error, allerror, indiroff, waitforupdate;
off_t osize;
ip = VTOI(vp);
ump = VFSTOUFS(vp->v_mount);
fs = ump->um_fs;
bo = &vp->v_bufobj;
ASSERT_VOP_LOCKED(vp, "ffs_truncate");
if (length < 0)
return (EINVAL);
if (length > fs->fs_maxfilesize)
return (EFBIG);
#ifdef QUOTA
error = getinoquota(ip);
if (error)
return (error);
#endif
/*
* Historically clients did not have to specify which data
* they were truncating. So, if not specified, we assume
* traditional behavior, e.g., just the normal data.
*/
if ((flags & (IO_EXT | IO_NORMAL)) == 0)
flags |= IO_NORMAL;
if (!DOINGSOFTDEP(vp) && !DOINGASYNC(vp))
flags |= IO_SYNC;
waitforupdate = (flags & IO_SYNC) != 0 || !DOINGASYNC(vp);
/*
* If we are truncating the extended-attributes, and cannot
* do it with soft updates, then do it slowly here. If we are
* truncating both the extended attributes and the file contents
* (e.g., the file is being unlinked), then pick it off with
* soft updates below.
*/
allerror = 0;
needextclean = 0;
softdeptrunc = 0;
journaltrunc = DOINGSUJ(vp);
if (journaltrunc == 0 && DOINGSOFTDEP(vp) && length == 0)
softdeptrunc = !softdep_slowdown(vp);
extblocks = 0;
datablocks = DIP(ip, i_blocks);
if (fs->fs_magic == FS_UFS2_MAGIC && ip->i_din2->di_extsize > 0) {
extblocks = btodb(fragroundup(fs, ip->i_din2->di_extsize));
datablocks -= extblocks;
}
if ((flags & IO_EXT) && extblocks > 0) {
if (length != 0)
panic("ffs_truncate: partial trunc of extdata");
if (softdeptrunc || journaltrunc) {
if ((flags & IO_NORMAL) == 0)
goto extclean;
needextclean = 1;
} else {
if ((error = ffs_syncvnode(vp, MNT_WAIT, 0)) != 0)
return (error);
#ifdef QUOTA
(void) chkdq(ip, -extblocks, NOCRED, 0);
#endif
vinvalbuf(vp, V_ALT, 0, 0);
vn_pages_remove(vp,
OFF_TO_IDX(lblktosize(fs, -extblocks)), 0);
osize = ip->i_din2->di_extsize;
ip->i_din2->di_blocks -= extblocks;
ip->i_din2->di_extsize = 0;
for (i = 0; i < UFS_NXADDR; i++) {
oldblks[i] = ip->i_din2->di_extb[i];
ip->i_din2->di_extb[i] = 0;
}
ip->i_flag |= IN_CHANGE;
if ((error = ffs_update(vp, waitforupdate)))
return (error);
for (i = 0; i < UFS_NXADDR; i++) {
if (oldblks[i] == 0)
continue;
ffs_blkfree(ump, fs, ITODEVVP(ip), oldblks[i],
sblksize(fs, osize, i), ip->i_number,
vp->v_type, nil);
}
}
}
if ((flags & IO_NORMAL) == 0)
return (0);
if (vp->v_type == VLNK &&
(ip->i_size < vp->v_mount->mnt_maxsymlinklen ||
datablocks == 0)) {
#ifdef INVARIANTS
if (length != 0)
panic("ffs_truncate: partial truncate of symlink");
#endif
bzero(SHORTLINK(ip), (uint)ip->i_size);
ip->i_size = 0;
DIP_SET(ip, i_size, 0);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
if (needextclean)
goto extclean;
return (ffs_update(vp, waitforupdate));
}
if (ip->i_size == length) {
ip->i_flag |= IN_CHANGE | IN_UPDATE;
if (needextclean)
goto extclean;
return (ffs_update(vp, 0));
}
if (fs->fs_ronly)
panic("ffs_truncate: read-only filesystem");
if (IS_SNAPSHOT(ip))
ffs_snapremove(vp);
vp->v_lasta = vp->v_clen = vp->v_cstart = vp->v_lastw = 0;
osize = ip->i_size;
/*
* Lengthen the size of the file. We must ensure that the
* last byte of the file is allocated. Since the smallest
* value of osize is 0, length will be at least 1.
*/
if (osize < length) {
vnode_pager_setsize(vp, length);
flags |= BA_CLRBUF;
error = UFS_BALLOC(vp, length - 1, 1, cred, flags, &bp);
if (error) {
vnode_pager_setsize(vp, osize);
return (error);
}
ip->i_size = length;
DIP_SET(ip, i_size, length);
if (bp->b_bufsize == fs->fs_bsize)
bp->b_flags |= B_CLUSTEROK;
if (flags & IO_SYNC)
bwrite(bp);
else if (DOINGASYNC(vp))
bdwrite(bp);
else
bawrite(bp);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
return (ffs_update(vp, waitforupdate));
}
/*
* Lookup block number for a given offset. Zero length files
* have no blocks, so return a blkno of -1.
*/
lbn = lblkno(fs, length - 1);
if (length == 0) {
blkno = -1;
} else if (lbn < UFS_NDADDR) {
blkno = DIP(ip, i_db[lbn]);
} else {
error = UFS_BALLOC(vp, lblktosize(fs, (off_t)lbn), fs->fs_bsize,
cred, BA_METAONLY, &bp);
if (error)
return (error);
indiroff = (lbn - UFS_NDADDR) % NINDIR(fs);
if (I_IS_UFS1(ip))
blkno = ((ufs1_daddr_t *)(bp->b_data))[indiroff];
else
blkno = ((ufs2_daddr_t *)(bp->b_data))[indiroff];
/*
* If the block number is non-zero, then the indirect block
* must have been previously allocated and need not be written.
* If the block number is zero, then we may have allocated
* the indirect block and hence need to write it out.
*/
if (blkno != 0)
brelse(bp);
else if (flags & IO_SYNC)
bwrite(bp);
else
bdwrite(bp);
}
/*
* If the block number at the new end of the file is zero,
* then we must allocate it to ensure that the last block of
* the file is allocated. Soft updates does not handle this
* case, so here we have to clean up the soft updates data
* structures describing the allocation past the truncation
* point. Finding and deallocating those structures is a lot of
* work. Since partial truncation with a hole at the end occurs
* rarely, we solve the problem by syncing the file so that it
* will have no soft updates data structures left.
*/
if (blkno == 0 && (error = ffs_syncvnode(vp, MNT_WAIT, 0)) != 0)
return (error);
if (blkno != 0 && DOINGSOFTDEP(vp)) {
if (softdeptrunc == 0 && journaltrunc == 0) {
/*
* If soft updates cannot handle this truncation,
* clean up soft dependency data structures and
* fall through to the synchronous truncation.
*/
if ((error = ffs_syncvnode(vp, MNT_WAIT, 0)) != 0)
return (error);
} else {
flags = IO_NORMAL | (needextclean ? IO_EXT: 0);
if (journaltrunc)
softdep_journal_freeblocks(ip, cred, length,
flags);
else
softdep_setup_freeblocks(ip, length, flags);
ASSERT_VOP_LOCKED(vp, "ffs_truncate1");
if (journaltrunc == 0) {
ip->i_flag |= IN_CHANGE | IN_UPDATE;
error = ffs_update(vp, 0);
}
return (error);
}
}
/*
* Shorten the size of the file. If the last block of the
* shortened file is unallocated, we must allocate it.
* Additionally, if the file is not being truncated to a
* block boundary, the contents of the partial block
* following the end of the file must be zero'ed in
* case it ever becomes accessible again because of
* subsequent file growth. Directories however are not
* zero'ed as they should grow back initialized to empty.
*/
offset = blkoff(fs, length);
if (blkno != 0 && offset == 0) {
ip->i_size = length;
DIP_SET(ip, i_size, length);
} else {
lbn = lblkno(fs, length);
flags |= BA_CLRBUF;
error = UFS_BALLOC(vp, length - 1, 1, cred, flags, &bp);
if (error)
return (error);
/*
* When we are doing soft updates and the UFS_BALLOC
* above fills in a direct block hole with a full sized
* block that will be truncated down to a fragment below,
* we must flush out the block dependency with an FSYNC
* so that we do not get a soft updates inconsistency
* when we create the fragment below.
*/
if (DOINGSOFTDEP(vp) && lbn < UFS_NDADDR &&
fragroundup(fs, blkoff(fs, length)) < fs->fs_bsize &&
(error = ffs_syncvnode(vp, MNT_WAIT, 0)) != 0)
return (error);
ip->i_size = length;
DIP_SET(ip, i_size, length);
size = blksize(fs, ip, lbn);
if (vp->v_type != VDIR && offset != 0)
bzero((char *)bp->b_data + offset,
(uint)(size - offset));
/* Kirk's code has reallocbuf(bp, size, 1) here */
allocbuf(bp, size);
if (bp->b_bufsize == fs->fs_bsize)
bp->b_flags |= B_CLUSTEROK;
if (flags & IO_SYNC)
bwrite(bp);
else if (DOINGASYNC(vp))
bdwrite(bp);
else
bawrite(bp);
}
/*
* Calculate index into inode's block list of
* last direct and indirect blocks (if any)
* which we want to keep. Lastblock is -1 when
* the file is truncated to 0.
*/
lastblock = lblkno(fs, length + fs->fs_bsize - 1) - 1;
lastiblock[SINGLE] = lastblock - UFS_NDADDR;
lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
nblocks = btodb(fs->fs_bsize);
/*
* Update file and block pointers on disk before we start freeing
* blocks. If we crash before free'ing blocks below, the blocks
* will be returned to the free list. lastiblock values are also
* normalized to -1 for calls to ffs_indirtrunc below.
*/
for (level = TRIPLE; level >= SINGLE; level--) {
oldblks[UFS_NDADDR + level] = DIP(ip, i_ib[level]);
if (lastiblock[level] < 0) {
DIP_SET(ip, i_ib[level], 0);
lastiblock[level] = -1;
}
}
for (i = 0; i < UFS_NDADDR; i++) {
oldblks[i] = DIP(ip, i_db[i]);
if (i > lastblock)
DIP_SET(ip, i_db[i], 0);
}
ip->i_flag |= IN_CHANGE | IN_UPDATE;
allerror = ffs_update(vp, waitforupdate);
/*
* Having written the new inode to disk, save its new configuration
* and put back the old block pointers long enough to process them.
* Note that we save the new block configuration so we can check it
* when we are done.
*/
for (i = 0; i < UFS_NDADDR; i++) {
newblks[i] = DIP(ip, i_db[i]);
DIP_SET(ip, i_db[i], oldblks[i]);
}
for (i = 0; i < UFS_NIADDR; i++) {
newblks[UFS_NDADDR + i] = DIP(ip, i_ib[i]);
DIP_SET(ip, i_ib[i], oldblks[UFS_NDADDR + i]);
}
ip->i_size = osize;
DIP_SET(ip, i_size, osize);
error = vtruncbuf(vp, cred, length, fs->fs_bsize);
if (error && (allerror == 0))
allerror = error;
/*
* Indirect blocks first.
*/
indir_lbn[SINGLE] = -UFS_NDADDR;
indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1;
indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
for (level = TRIPLE; level >= SINGLE; level--) {
bn = DIP(ip, i_ib[level]);
if (bn != 0) {
error = ffs_indirtrunc(ip, indir_lbn[level],
fsbtodb(fs, bn), lastiblock[level], level, &count);
if (error)
allerror = error;
blocksreleased += count;
if (lastiblock[level] < 0) {
DIP_SET(ip, i_ib[level], 0);
ffs_blkfree(ump, fs, ump->um_devvp, bn,
fs->fs_bsize, ip->i_number,
vp->v_type, nil);
blocksreleased += nblocks;
}
}
if (lastiblock[level] >= 0)
goto done;
}
/*
* All whole direct blocks or frags.
*/
for (i = UFS_NDADDR - 1; i > lastblock; i--) {
long bsize;
bn = DIP(ip, i_db[i]);
if (bn == 0)
continue;
DIP_SET(ip, i_db[i], 0);
bsize = blksize(fs, ip, i);
ffs_blkfree(ump, fs, ump->um_devvp, bn, bsize, ip->i_number,
vp->v_type, nil);
blocksreleased += btodb(bsize);
}
if (lastblock < 0)
goto done;
/*
* Finally, look for a change in size of the
* last direct block; release any frags.
*/
bn = DIP(ip, i_db[lastblock]);
if (bn != 0) {
long oldspace, newspace;
/*
* Calculate amount of space we're giving
* back as old block size minus new block size.
*/
oldspace = blksize(fs, ip, lastblock);
ip->i_size = length;
DIP_SET(ip, i_size, length);
newspace = blksize(fs, ip, lastblock);
if (newspace == 0)
panic("ffs_truncate: newspace");
if (oldspace - newspace > 0) {
/*
* Block number of space to be free'd is
* the old block # plus the number of frags
* required for the storage we're keeping.
*/
bn += numfrags(fs, newspace);
ffs_blkfree(ump, fs, ump->um_devvp, bn,
oldspace - newspace, ip->i_number, vp->v_type, nil);
blocksreleased += btodb(oldspace - newspace);
}
}
done:
#ifdef INVARIANTS
for (level = SINGLE; level <= TRIPLE; level++)
if (newblks[UFS_NDADDR + level] != DIP(ip, i_ib[level]))
panic("ffs_truncate1");
for (i = 0; i < UFS_NDADDR; i++)
if (newblks[i] != DIP(ip, i_db[i]))
panic("ffs_truncate2");
BO_LOCK(bo);
if (length == 0 &&
(fs->fs_magic != FS_UFS2_MAGIC || ip->i_din2->di_extsize == 0) &&
(bo->bo_dirty.bv_cnt > 0 || bo->bo_clean.bv_cnt > 0))
panic("ffs_truncate3");
BO_UNLOCK(bo);
#endif /* INVARIANTS */
/*
* Put back the real size.
*/
ip->i_size = length;
DIP_SET(ip, i_size, length);
if (DIP(ip, i_blocks) >= blocksreleased)
DIP_SET(ip, i_blocks, DIP(ip, i_blocks) - blocksreleased);
else /* sanity */
DIP_SET(ip, i_blocks, 0);
ip->i_flag |= IN_CHANGE;
#ifdef QUOTA
(void) chkdq(ip, -blocksreleased, NOCRED, 0);
#endif
return (allerror);
extclean:
if (journaltrunc)
softdep_journal_freeblocks(ip, cred, length, IO_EXT);
else
softdep_setup_freeblocks(ip, length, IO_EXT);
return (ffs_update(vp, waitforupdate));
#endif // 0
return 0;
}