static ino_t journal_findfile( void ) {
struct ufs1_dinode *dp1;
struct ufs2_dinode *dp2;
ino_t ino;
int mode;
void *ip;
int i;
if ( getino( &disk, &ip, ROOTINO, &mode ) != 0 ) {
warn( "Failed to get root inode" );
return ( -1 );
}
dp2 = ip;
dp1 = ip;
if ( sblock.fs_magic == FS_UFS1_MAGIC ) {
if ( (off_t) dp1->di_size >= lblktosize( &sblock, NDADDR ) ) {
warnx( "ROOTINO extends beyond direct blocks." );
return ( -1 );
}
for ( i = 0; i < NDADDR; i++ ) {
if ( dp1->di_db[i] == 0 ) break;
if ( ( ino = dir_search( dp1->di_db[i], sblksize( &sblock, (off_t) dp1->di_size, i ) ) ) != 0 ) return ( ino );
}
} else {
if ( (off_t) dp2->di_size >= lblktosize( &sblock, NDADDR ) ) {
warnx( "ROOTINO extends beyond direct blocks." );
return ( -1 );
}
for ( i = 0; i < NDADDR; i++ ) {
if ( dp2->di_db[i] == 0 ) break;
if ( ( ino = dir_search( dp2->di_db[i], sblksize( &sblock, (off_t) dp2->di_size, i ) ) ) != 0 ) return ( ino );
}
}
return ( 0 );
}
int
wapbl_allocate_log_file(struct mount *mp, struct vnode *vp,
daddr_t *startp, size_t *countp, uint64_t *extradatap)
{
struct ufsmount *ump = VFSTOUFS(mp);
struct fs *fs = ump->um_fs;
struct inode *ip = VTOI(vp);
daddr_t addr, indir_addr;
off_t logsize;
size_t size;
int error;
logsize = 0;
/* check if there's a suggested log size */
if (fs->fs_journal_flags & UFS_WAPBL_FLAGS_CREATE_LOG &&
fs->fs_journal_location == UFS_WAPBL_JOURNALLOC_IN_FILESYSTEM)
logsize = fs->fs_journallocs[UFS_WAPBL_INFS_COUNT];
if (DIP(ip, size) > 0) {
printf("%s: file size (%lld) non zero\n", __func__,
DIP(ip, size));
return EEXIST;
}
wapbl_find_log_start(mp, vp, logsize, &addr, &indir_addr, &size);
if (addr == 0) {
printf("%s: log not allocated, largest extent is "
"%lldMB\n", __func__,
lblktosize(fs, size) / (1024 * 1024));
return ENOSPC;
}
logsize = lblktosize(fs, size); /* final log size */
VTOI(vp)->i_ffs_first_data_blk = addr;
VTOI(vp)->i_ffs_first_indir_blk = indir_addr;
error = ufs_gop_alloc(vp, 0, logsize, B_CONTIG, curproc->p_ucred);
if (error) {
printf("%s: ufs_gop_alloc error %d\n", __func__, error);
return error;
}
*startp = fsbtodb(fs, addr);
*countp = btodb(logsize);
*extradatap = VTOI(vp)->i_number;
return 0;
}
static int
udf_read(struct vop_read_args *ap)
{
struct vnode *vp = ap->a_vp;
struct uio *uio = ap->a_uio;
struct udf_node *node = VTON(vp);
struct udf_mnt *udfmp;
struct file_entry *fentry;
struct buf *bp;
uint8_t *data;
daddr_t lbn, rablock;
off_t diff, fsize;
ssize_t n;
int error = 0;
long size, on;
if (uio->uio_resid == 0)
return (0);
if (uio->uio_offset < 0)
return (EINVAL);
if (is_data_in_fentry(node)) {
fentry = node->fentry;
data = &fentry->data[le32toh(fentry->l_ea)];
fsize = le32toh(fentry->l_ad);
n = uio->uio_resid;
diff = fsize - uio->uio_offset;
if (diff <= 0)
return (0);
if (diff < n)
n = diff;
error = uiomove(data + uio->uio_offset, (int)n, uio);
return (error);
}
fsize = le64toh(node->fentry->inf_len);
udfmp = node->udfmp;
do {
lbn = lblkno(udfmp, uio->uio_offset);
on = blkoff(udfmp, uio->uio_offset);
n = min((u_int)(udfmp->bsize - on),
uio->uio_resid);
diff = fsize - uio->uio_offset;
if (diff <= 0)
return (0);
if (diff < n)
n = diff;
size = udfmp->bsize;
rablock = lbn + 1;
if ((vp->v_mount->mnt_flag & MNT_NOCLUSTERR) == 0) {
if (lblktosize(udfmp, rablock) < fsize) {
error = cluster_read(vp, fsize, lbn, size,
NOCRED, uio->uio_resid,
(ap->a_ioflag >> 16), 0, &bp);
} else {
error = bread(vp, lbn, size, NOCRED, &bp);
}
} else {
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
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);
}
/*
* Insert the journal file into the ROOTINO directory. We always extend the
* last frag
*/
static int
journal_insertfile(ino_t ino)
{
struct ufs1_dinode *dp1;
struct ufs2_dinode *dp2;
void *ip;
ufs2_daddr_t nblk;
ufs2_daddr_t blk;
ufs_lbn_t lbn;
int size;
int mode;
int off;
if (getino(&disk, &ip, ROOTINO, &mode) != 0) {
warn("Failed to get root inode");
sbdirty();
return (-1);
}
dp2 = ip;
dp1 = ip;
blk = 0;
size = 0;
nblk = journal_balloc();
if (nblk <= 0)
return (-1);
/*
* For simplicity sake we aways extend the ROOTINO into a new
* directory block rather than searching for space and inserting
* into an existing block. However, if the rootino has frags
* have to free them and extend the block.
*/
if (sblock.fs_magic == FS_UFS1_MAGIC) {
lbn = lblkno(&sblock, dp1->di_size);
off = blkoff(&sblock, dp1->di_size);
blk = dp1->di_db[lbn];
size = sblksize(&sblock, (off_t)dp1->di_size, lbn);
} else {
lbn = lblkno(&sblock, dp2->di_size);
off = blkoff(&sblock, dp2->di_size);
blk = dp2->di_db[lbn];
size = sblksize(&sblock, (off_t)dp2->di_size, lbn);
}
if (off != 0) {
if (dir_extend(blk, nblk, off, ino) == -1)
return (-1);
} else {
blk = 0;
if (dir_insert(nblk, 0, ino) == -1)
return (-1);
}
if (sblock.fs_magic == FS_UFS1_MAGIC) {
dp1->di_blocks += (sblock.fs_bsize - size) / DEV_BSIZE;
dp1->di_db[lbn] = nblk;
dp1->di_size = lblktosize(&sblock, lbn+1);
} else {
dp2->di_blocks += (sblock.fs_bsize - size) / DEV_BSIZE;
dp2->di_db[lbn] = nblk;
dp2->di_size = lblktosize(&sblock, lbn+1);
}
if (putino(&disk) < 0) {
warn("Failed to write root inode");
return (-1);
}
if (cgwrite(&disk) < 0) {
warn("Failed to write updated cg");
sbdirty();
return (-1);
}
if (blk) {
if (cgbfree(&disk, blk, size) < 0) {
warn("Failed to write cg");
return (-1);
}
}
return (0);
}
/*
* Find a suitable location for the journal in the filesystem.
*
* Our strategy here is to look for a contiguous block of free space
* at least "logfile" MB in size (plus room for any indirect blocks).
* We start at the middle of the filesystem and check each cylinder
* group working outwards. If "logfile" MB is not available as a
* single contigous chunk, then return the address and size of the
* largest chunk found.
*
* XXX
* At what stage does the search fail? Is if the largest space we could
* find is less than a quarter the requested space reasonable? If the
* search fails entirely, return a block address if "0" it indicate this.
*/
void
wapbl_find_log_start(struct mount *mp, struct vnode *vp, off_t logsize,
daddr_t *addr, daddr_t *indir_addr, size_t *size)
{
struct ufsmount *ump = VFSTOUFS(mp);
struct fs *fs = ump->um_fs;
struct vnode *devvp = ump->um_devvp;
struct cg *cgp;
struct buf *bp;
uint8_t *blksfree;
daddr_t blkno, best_addr, start_addr;
daddr_t desired_blks, min_desired_blks;
daddr_t freeblks, best_blks;
int bpcg, cg, error, fixedsize, indir_blks, n, s;
#ifdef FFS_EI
const int needswap = UFS_FSNEEDSWAP(fs);
#endif
if (logsize == 0) {
fixedsize = 0; /* We can adjust the size if tight */
logsize = lfragtosize(fs, fs->fs_dsize) /
UFS_WAPBL_JOURNAL_SCALE;
DPRINTF("suggested log size = %lld\n", logsize);
logsize = max(logsize, UFS_WAPBL_MIN_JOURNAL_SIZE);
logsize = min(logsize, UFS_WAPBL_MAX_JOURNAL_SIZE);
DPRINTF("adjusted log size = %lld\n", logsize);
} else {
fixedsize = 1;
DPRINTF("fixed log size = %lld\n", logsize);
}
desired_blks = logsize / fs->fs_bsize;
DPRINTF("desired blocks = %lld\n", desired_blks);
/* add in number of indirect blocks needed */
indir_blks = 0;
if (desired_blks >= NDADDR) {
struct indir indirs[NIADDR + 2];
int num;
error = ufs_getlbns(vp, desired_blks, indirs, &num);
if (error) {
printf("%s: ufs_getlbns failed, error %d!\n",
__func__, error);
goto bad;
}
switch (num) {
case 2:
indir_blks = 1; /* 1st level indirect */
break;
case 3:
indir_blks = 1 + /* 1st level indirect */
1 + /* 2nd level indirect */
indirs[1].in_off + 1; /* extra 1st level indirect */
break;
default:
printf("%s: unexpected numlevels %d from ufs_getlbns\n",
__func__, num);
*size = 0;
goto bad;
}
desired_blks += indir_blks;
}
DPRINTF("desired blocks = %lld (including indirect)\n",
desired_blks);
/*
* If a specific size wasn't requested, allow for a smaller log
* if we're really tight for space...
*/
min_desired_blks = desired_blks;
if (!fixedsize)
min_desired_blks = desired_blks / 4;
/* Look at number of blocks per CG. If it's too small, bail early. */
bpcg = fragstoblks(fs, fs->fs_fpg);
if (min_desired_blks > bpcg) {
printf("ffs_wapbl: cylinder group size of %lld MB "
" is not big enough for journal\n",
lblktosize(fs, bpcg) / (1024 * 1024));
goto bad;
}
/*
* Start with the middle cylinder group, and search outwards in
* both directions until we either find the requested log size
* or reach the start/end of the file system. If we reach the
* start/end without finding enough space for the full requested
* log size, use the largest extent found if it is large enough
* to satisfy the our minimum size.
*
* XXX
* Can we just use the cluster contigsum stuff (esp on UFS2)
* here to simplify this search code?
*/
best_addr = 0;
best_blks = 0;
for (cg = fs->fs_ncg / 2, s = 0, n = 1;
best_blks < desired_blks && cg >= 0 && cg < fs->fs_ncg;
s++, n = -n, cg += n * s) {
DPRINTF("check cg %d of %d\n", cg, fs->fs_ncg);
error = bread(devvp, fsbtodb(fs, cgtod(fs, cg)),
fs->fs_cgsize, &bp);
if (error) {
continue;
}
cgp = (struct cg *)bp->b_data;
if (!cg_chkmagic(cgp)) {
brelse(bp);
continue;
}
blksfree = cg_blksfree(cgp);
for (blkno = 0; blkno < bpcg;) {
/* look for next free block */
/* XXX use scanc() and fragtbl[] here? */
for (; blkno < bpcg - min_desired_blks; blkno++)
if (ffs_isblock(fs, blksfree, blkno))
break;
/* past end of search space in this CG? */
if (blkno >= bpcg - min_desired_blks)
break;
/* count how many free blocks in this extent */
start_addr = blkno;
for (freeblks = 0; blkno < bpcg; blkno++, freeblks++)
if (!ffs_isblock(fs, blksfree, blkno))
break;
if (freeblks > best_blks) {
best_blks = freeblks;
best_addr = blkstofrags(fs, start_addr) +
cgbase(fs, cg);
if (freeblks >= desired_blks) {
DPRINTF("found len %lld"
" at offset %lld in gc\n",
freeblks, start_addr);
break;
}
}
}
brelse(bp);
}
DPRINTF("best found len = %lld, wanted %lld"
" at addr %lld\n", best_blks, desired_blks, best_addr);
if (best_blks < min_desired_blks) {
*addr = 0;
*indir_addr = 0;
} else {
/* put indirect blocks at start, and data blocks after */
*addr = best_addr + blkstofrags(fs, indir_blks);
*indir_addr = best_addr;
}
*size = min(desired_blks, best_blks) - indir_blks;
return;
bad:
*addr = 0;
*indir_addr = 0;
*size = 0;
return;
}
/* ARGSUSED */
int
ext2fs_read(void *v)
{
struct vop_read_args *ap = v;
struct vnode *vp;
struct inode *ip;
struct uio *uio;
struct m_ext2fs *fs;
struct buf *bp;
daddr_t lbn, nextlbn;
off_t bytesinfile;
long size, xfersize, blkoffset;
int error;
vp = ap->a_vp;
ip = VTOI(vp);
uio = ap->a_uio;
#ifdef DIAGNOSTIC
if (uio->uio_rw != UIO_READ)
panic("%s: mode", "ext2fs_read");
if (vp->v_type == VLNK) {
if ((int)ext2fs_size(ip) < vp->v_mount->mnt_maxsymlinklen ||
(vp->v_mount->mnt_maxsymlinklen == 0 &&
ip->i_e2fs_nblock == 0))
panic("%s: short symlink", "ext2fs_read");
} else if (vp->v_type != VREG && vp->v_type != VDIR)
panic("%s: type %d", "ext2fs_read", vp->v_type);
#endif
fs = ip->i_e2fs;
if ((u_int64_t)uio->uio_offset >
((u_int64_t)0x80000000 * fs->e2fs_bsize - 1))
return (EFBIG);
if (uio->uio_resid == 0)
return (0);
for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) {
if ((bytesinfile = ext2fs_size(ip) - uio->uio_offset) <= 0)
break;
lbn = lblkno(fs, uio->uio_offset);
nextlbn = lbn + 1;
size = fs->e2fs_bsize;
blkoffset = blkoff(fs, uio->uio_offset);
xfersize = fs->e2fs_bsize - blkoffset;
if (uio->uio_resid < xfersize)
xfersize = uio->uio_resid;
if (bytesinfile < xfersize)
xfersize = bytesinfile;
if (lblktosize(fs, nextlbn) >= ext2fs_size(ip))
error = bread(vp, lbn, size, &bp);
else if (lbn - 1 == ip->i_ci.ci_lastr) {
int nextsize = fs->e2fs_bsize;
error = breadn(vp, lbn, size, &nextlbn, &nextsize,
1, &bp);
} else
error = bread(vp, lbn, size, &bp);
if (error)
break;
ip->i_ci.ci_lastr = lbn;
/*
* We should only get non-zero b_resid when an I/O error
* has occurred, which should cause us to break above.
* However, if the short read did not cause an error,
* then we want to ensure that we do not uiomove bad
* or uninitialized data.
*/
size -= bp->b_resid;
if (size < xfersize) {
if (size == 0)
break;
xfersize = size;
}
error = uiomove((char *)bp->b_data + blkoffset, xfersize, uio);
if (error)
break;
brelse(bp);
}
if (bp != NULL)
brelse(bp);
if (!(vp->v_mount->mnt_flag & MNT_NOATIME)) {
ip->i_flag |= IN_ACCESS;
}
return (error);
}
/*
* Truncate the inode oip to at most length size, freeing the
* disk blocks.
*/
int
ffs_truncate(struct vnode *ovp, off_t length, int ioflag, kauth_cred_t cred)
{
daddr_t lastblock;
struct inode *oip = VTOI(ovp);
daddr_t bn, lastiblock[NIADDR], indir_lbn[NIADDR];
daddr_t blks[NDADDR + NIADDR];
struct fs *fs;
int offset, pgoffset, level;
int64_t count, blocksreleased = 0;
int i, aflag, nblocks;
int error, allerror = 0;
off_t osize;
int sync;
struct ufsmount *ump = oip->i_ump;
if (ovp->v_type == VCHR || ovp->v_type == VBLK ||
ovp->v_type == VFIFO || ovp->v_type == VSOCK) {
KASSERT(oip->i_size == 0);
return 0;
}
if (length < 0)
return (EINVAL);
if (ovp->v_type == VLNK &&
(oip->i_size < ump->um_maxsymlinklen ||
(ump->um_maxsymlinklen == 0 && DIP(oip, blocks) == 0))) {
KDASSERT(length == 0);
memset(SHORTLINK(oip), 0, (size_t)oip->i_size);
oip->i_size = 0;
DIP_ASSIGN(oip, size, 0);
oip->i_flag |= IN_CHANGE | IN_UPDATE;
return (ffs_update(ovp, NULL, NULL, 0));
}
if (oip->i_size == length) {
/* still do a uvm_vnp_setsize() as writesize may be larger */
uvm_vnp_setsize(ovp, length);
oip->i_flag |= IN_CHANGE | IN_UPDATE;
return (ffs_update(ovp, NULL, NULL, 0));
}
fs = oip->i_fs;
if (length > ump->um_maxfilesize)
return (EFBIG);
if ((oip->i_flags & SF_SNAPSHOT) != 0)
ffs_snapremove(ovp);
osize = oip->i_size;
aflag = ioflag & IO_SYNC ? B_SYNC : 0;
/*
* 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 (lblkno(fs, osize) < NDADDR &&
lblkno(fs, osize) != lblkno(fs, length) &&
blkroundup(fs, osize) != osize) {
off_t eob;
eob = blkroundup(fs, osize);
uvm_vnp_setwritesize(ovp, eob);
error = ufs_balloc_range(ovp, osize, eob - osize,
cred, aflag);
if (error) {
(void) ffs_truncate(ovp, osize,
ioflag & IO_SYNC, cred);
return error;
}
if (ioflag & IO_SYNC) {
mutex_enter(ovp->v_interlock);
VOP_PUTPAGES(ovp,
trunc_page(osize & fs->fs_bmask),
round_page(eob), PGO_CLEANIT | PGO_SYNCIO |
PGO_JOURNALLOCKED);
}
}
uvm_vnp_setwritesize(ovp, length);
error = ufs_balloc_range(ovp, length - 1, 1, cred, aflag);
if (error) {
(void) ffs_truncate(ovp, osize, ioflag & IO_SYNC, cred);
return (error);
}
uvm_vnp_setsize(ovp, length);
oip->i_flag |= IN_CHANGE | IN_UPDATE;
KASSERT(ovp->v_size == oip->i_size);
return (ffs_update(ovp, NULL, NULL, 0));
}
/*
* When truncating a regular file down to a non-block-aligned size,
* we must zero the part of last block which is past the new EOF.
* We must synchronously flush the zeroed pages to disk
* since the new pages will be invalidated as soon as we
* inform the VM system of the new, smaller size.
* We must do this before acquiring the GLOCK, since fetching
* the pages will acquire the GLOCK internally.
* So there is a window where another thread could see a whole
* zeroed page past EOF, but that's life.
*/
offset = blkoff(fs, length);
pgoffset = length & PAGE_MASK;
if (ovp->v_type == VREG && (pgoffset != 0 || offset != 0) &&
osize > length) {
daddr_t lbn;
voff_t eoz;
int size;
if (offset != 0) {
error = ufs_balloc_range(ovp, length - 1, 1, cred,
aflag);
if (error)
return error;
}
lbn = lblkno(fs, length);
size = blksize(fs, oip, lbn);
eoz = MIN(MAX(lblktosize(fs, lbn) + size, round_page(pgoffset)),
osize);
ubc_zerorange(&ovp->v_uobj, length, eoz - length,
UBC_UNMAP_FLAG(ovp));
if (round_page(eoz) > round_page(length)) {
mutex_enter(ovp->v_interlock);
error = VOP_PUTPAGES(ovp, round_page(length),
round_page(eoz),
PGO_CLEANIT | PGO_DEACTIVATE | PGO_JOURNALLOCKED |
((ioflag & IO_SYNC) ? PGO_SYNCIO : 0));
if (error)
return error;
}
}
genfs_node_wrlock(ovp);
oip->i_size = length;
DIP_ASSIGN(oip, size, length);
uvm_vnp_setsize(ovp, length);
/*
* 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.
*/
sync = 0;
for (level = TRIPLE; level >= SINGLE; level--) {
blks[NDADDR + level] = DIP(oip, ib[level]);
if (lastiblock[level] < 0 && blks[NDADDR + level] != 0) {
sync = 1;
DIP_ASSIGN(oip, ib[level], 0);
lastiblock[level] = -1;
}
}
for (i = 0; i < NDADDR; i++) {
blks[i] = DIP(oip, db[i]);
if (i > lastblock && blks[i] != 0) {
sync = 1;
DIP_ASSIGN(oip, db[i], 0);
}
}
oip->i_flag |= IN_CHANGE | IN_UPDATE;
if (sync) {
error = ffs_update(ovp, NULL, NULL, UPDATE_WAIT);
if (error && !allerror)
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++) {
bn = DIP(oip, db[i]);
DIP_ASSIGN(oip, db[i], blks[i]);
blks[i] = bn;
}
for (i = 0; i < NIADDR; i++) {
bn = DIP(oip, ib[i]);
DIP_ASSIGN(oip, ib[i], blks[NDADDR + i]);
blks[NDADDR + i] = bn;
}
oip->i_size = osize;
DIP_ASSIGN(oip, size, osize);
error = vtruncbuf(ovp, lastblock + 1, 0, 0);
if (error && !allerror)
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--) {
if (oip->i_ump->um_fstype == UFS1)
bn = ufs_rw32(oip->i_ffs1_ib[level],UFS_FSNEEDSWAP(fs));
else
bn = ufs_rw64(oip->i_ffs2_ib[level],UFS_FSNEEDSWAP(fs));
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);
if (oip->i_ump->um_mountp->mnt_wapbl) {
UFS_WAPBL_REGISTER_DEALLOCATION(
oip->i_ump->um_mountp,
fsbtodb(fs, bn), fs->fs_bsize);
} else
ffs_blkfree(fs, oip->i_devvp, bn,
fs->fs_bsize, oip->i_number);
blocksreleased += nblocks;
}
}
if (lastiblock[level] >= 0)
goto done;
}
/*
* All whole direct blocks or frags.
*/
for (i = NDADDR - 1; i > lastblock; i--) {
long bsize;
if (oip->i_ump->um_fstype == UFS1)
bn = ufs_rw32(oip->i_ffs1_db[i], UFS_FSNEEDSWAP(fs));
else
bn = ufs_rw64(oip->i_ffs2_db[i], UFS_FSNEEDSWAP(fs));
if (bn == 0)
continue;
DIP_ASSIGN(oip, db[i], 0);
bsize = blksize(fs, oip, i);
if ((oip->i_ump->um_mountp->mnt_wapbl) &&
(ovp->v_type != VREG)) {
UFS_WAPBL_REGISTER_DEALLOCATION(oip->i_ump->um_mountp,
fsbtodb(fs, bn), bsize);
} else
ffs_blkfree(fs, oip->i_devvp, bn, bsize, oip->i_number);
blocksreleased += btodb(bsize);
}
if (lastblock < 0)
goto done;
/*
* Finally, look for a change in size of the
* last direct block; release any frags.
*/
if (oip->i_ump->um_fstype == UFS1)
bn = ufs_rw32(oip->i_ffs1_db[lastblock], UFS_FSNEEDSWAP(fs));
else
bn = ufs_rw64(oip->i_ffs2_db[lastblock], UFS_FSNEEDSWAP(fs));
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;
DIP_ASSIGN(oip, size, length);
newspace = blksize(fs, oip, lastblock);
if (newspace == 0)
panic("itrunc: 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);
if ((oip->i_ump->um_mountp->mnt_wapbl) &&
(ovp->v_type != VREG)) {
UFS_WAPBL_REGISTER_DEALLOCATION(
oip->i_ump->um_mountp, fsbtodb(fs, bn),
oldspace - newspace);
} else
ffs_blkfree(fs, oip->i_devvp, bn,
oldspace - newspace, oip->i_number);
blocksreleased += btodb(oldspace - newspace);
}
}
done:
#ifdef DIAGNOSTIC
for (level = SINGLE; level <= TRIPLE; level++)
if (blks[NDADDR + level] != DIP(oip, ib[level]))
panic("itrunc1");
for (i = 0; i < NDADDR; i++)
if (blks[i] != DIP(oip, db[i]))
panic("itrunc2");
if (length == 0 &&
(!LIST_EMPTY(&ovp->v_cleanblkhd) || !LIST_EMPTY(&ovp->v_dirtyblkhd)))
panic("itrunc3");
#endif /* DIAGNOSTIC */
/*
* Put back the real size.
*/
oip->i_size = length;
DIP_ASSIGN(oip, size, length);
DIP_ADD(oip, blocks, -blocksreleased);
genfs_node_unlock(ovp);
oip->i_flag |= IN_CHANGE;
UFS_WAPBL_UPDATE(ovp, NULL, NULL, 0);
#if defined(QUOTA) || defined(QUOTA2)
(void) chkdq(oip, -blocksreleased, NOCRED, 0);
#endif
KASSERT(ovp->v_type != VREG || ovp->v_size == oip->i_size);
return (allerror);
}
/*
* 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.
*/
int
ffs1_balloc(struct inode *ip, off_t startoffset, int size, struct ucred *cred,
int flags, struct buf **bpp)
{
daddr_t lbn, nb, newb, pref;
struct fs *fs;
struct buf *bp, *nbp;
struct vnode *vp;
struct proc *p;
struct indir indirs[NIADDR + 2];
int32_t *bap;
int deallocated, osize, nsize, num, i, error;
int32_t *allocib, *blkp, *allocblk, allociblk[NIADDR+1];
int unwindidx = -1;
vp = ITOV(ip);
fs = ip->i_fs;
p = curproc;
lbn = lblkno(fs, startoffset);
size = blkoff(fs, startoffset) + size;
if (size > fs->fs_bsize)
panic("ffs1_balloc: blk too big");
if (bpp != NULL)
*bpp = NULL;
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.
*/
nb = lblkno(fs, ip->i_ffs1_size);
if (nb < NDADDR && nb < lbn) {
osize = blksize(fs, ip, nb);
if (osize < fs->fs_bsize && osize > 0) {
error = ffs_realloccg(ip, nb,
ffs1_blkpref(ip, nb, (int)nb, &ip->i_ffs1_db[0]),
osize, (int)fs->fs_bsize, cred, bpp, &newb);
if (error)
return (error);
if (DOINGSOFTDEP(vp))
softdep_setup_allocdirect(ip, nb, newb,
ip->i_ffs1_db[nb], fs->fs_bsize, osize,
bpp ? *bpp : NULL);
ip->i_ffs1_size = lblktosize(fs, nb + 1);
uvm_vnp_setsize(vp, ip->i_ffs1_size);
ip->i_ffs1_db[nb] = newb;
ip->i_flag |= IN_CHANGE | IN_UPDATE;
if (bpp != NULL) {
if (flags & B_SYNC)
bwrite(*bpp);
else
bawrite(*bpp);
}
}
}
/*
* The first NDADDR blocks are direct blocks
*/
if (lbn < NDADDR) {
nb = ip->i_ffs1_db[lbn];
if (nb != 0 && ip->i_ffs1_size >= 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(vp, lbn, fs->fs_bsize, 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(vp, lbn, fs->fs_bsize,
bpp);
if (error) {
brelse(*bpp);
return (error);
}
(*bpp)->b_bcount = osize;
}
return (0);
} else {
/*
* The existing block is smaller than we
* want, grow it.
*/
error = ffs_realloccg(ip, lbn,
ffs1_blkpref(ip, lbn, (int)lbn,
&ip->i_ffs1_db[0]),
osize, nsize, cred, bpp, &newb);
if (error)
return (error);
if (DOINGSOFTDEP(vp))
softdep_setup_allocdirect(ip, lbn,
newb, nb, nsize, osize,
bpp ? *bpp : NULL);
}
} else {
/*
* The block was not previously allocated,
* allocate a new block or fragment.
*/
if (ip->i_ffs1_size < lblktosize(fs, lbn + 1))
nsize = fragroundup(fs, size);
else
nsize = fs->fs_bsize;
error = ffs_alloc(ip, lbn,
ffs1_blkpref(ip, lbn, (int)lbn, &ip->i_ffs1_db[0]),
nsize, cred, &newb);
if (error)
return (error);
if (bpp != NULL) {
*bpp = getblk(vp, lbn, fs->fs_bsize, 0, 0);
if (nsize < fs->fs_bsize)
(*bpp)->b_bcount = nsize;
(*bpp)->b_blkno = fsbtodb(fs, newb);
if (flags & B_CLRBUF)
clrbuf(*bpp);
}
if (DOINGSOFTDEP(vp))
softdep_setup_allocdirect(ip, lbn, newb, 0,
nsize, 0, bpp ? *bpp : NULL);
}
ip->i_ffs1_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)) != 0)
return(error);
#ifdef DIAGNOSTIC
if (num < 1)
panic ("ffs1_balloc: ufs_bmaparray returned indirect block");
#endif
/*
* Fetch the first indirect block allocating if necessary.
*/
--num;
nb = ip->i_ffs1_ib[indirs[0].in_off];
allocib = NULL;
allocblk = allociblk;
if (nb == 0) {
pref = ffs1_blkpref(ip, lbn, -indirs[0].in_off - 1, NULL);
error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize,
cred, &newb);
if (error)
goto fail;
nb = newb;
*allocblk++ = nb;
bp = getblk(vp, indirs[1].in_lbn, fs->fs_bsize, 0, 0);
bp->b_blkno = fsbtodb(fs, nb);
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 ((error = bwrite(bp)) != 0)
goto fail;
}
allocib = &ip->i_ffs1_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, indirs[i].in_lbn, (int)fs->fs_bsize, &bp);
if (error) {
brelse(bp);
goto fail;
}
bap = (int32_t *)bp->b_data;
nb = bap[indirs[i].in_off];
if (i == num)
break;
i++;
if (nb != 0) {
brelse(bp);
continue;
}
if (pref == 0)
pref = ffs1_blkpref(ip, lbn, i - num - 1, NULL);
error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, cred,
&newb);
if (error) {
brelse(bp);
goto fail;
}
nb = newb;
*allocblk++ = nb;
nbp = getblk(vp, indirs[i].in_lbn, fs->fs_bsize, 0, 0);
nbp->b_blkno = fsbtodb(fs, nb);
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 {
bdwrite(bp);
}
}
/*
* Get the data block, allocating if necessary.
*/
if (nb == 0) {
pref = ffs1_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;
if (bpp != NULL) {
nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0);
nbp->b_blkno = fsbtodb(fs, nb);
if (flags & B_CLRBUF)
clrbuf(nbp);
*bpp = nbp;
}
if (DOINGSOFTDEP(vp))
softdep_setup_allocindir_page(ip, lbn, bp,
indirs[i].in_off, nb, 0, bpp ? *bpp : NULL);
bap[indirs[i].in_off] = nb;
/*
* If required, write synchronously, otherwise use
* delayed write.
*/
if (flags & B_SYNC) {
bwrite(bp);
} else {
bdwrite(bp);
}
return (0);
}
brelse(bp);
if (bpp != NULL) {
if (flags & B_CLRBUF) {
error = bread(vp, lbn, (int)fs->fs_bsize, &nbp);
if (error) {
brelse(nbp);
goto fail;
}
} else {
nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0);
nbp->b_blkno = fsbtodb(fs, nb);
}
*bpp = nbp;
}
return (0);
fail:
/*
* If we have failed to allocate any blocks, simply return the error.
* This is the usual case and avoids the need to fsync the file.
*/
if (allocblk == allociblk && allocib == NULL && unwindidx == -1)
return (error);
/*
* If we have failed part way through block allocation, we have to
* deallocate any indirect blocks that we have allocated. We have to
* fsync the file before we start to get rid of all of its
* dependencies so that we do not leave them dangling. We have to sync
* it at the end so that the softdep code does not find any untracked
* changes. Although this is really slow, running out of disk space is
* not expected to be a common occurrence. The error return from fsync
* is ignored as we already have an error to return to the user.
*/
VOP_FSYNC(vp, p->p_ucred, MNT_WAIT, p);
for (deallocated = 0, blkp = allociblk; blkp < allocblk; blkp++) {
ffs_blkfree(ip, *blkp, fs->fs_bsize);
deallocated += fs->fs_bsize;
}
if (allocib != NULL) {
*allocib = 0;
} else if (unwindidx >= 0) {
int r;
r = bread(vp, indirs[unwindidx].in_lbn, (int)fs->fs_bsize, &bp);
if (r)
panic("Could not unwind indirect block, error %d", r);
bap = (int32_t *)bp->b_data;
bap[indirs[unwindidx].in_off] = 0;
if (flags & B_SYNC) {
bwrite(bp);
} else {
bdwrite(bp);
}
}
if (deallocated) {
/*
* Restore user's disk quota because allocation failed.
*/
(void)ufs_quota_free_blocks(ip, btodb(deallocated), cred);
ip->i_ffs1_blocks -= btodb(deallocated);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
}
VOP_FSYNC(vp, p->p_ucred, MNT_WAIT, p);
return (error);
}
/*
* Extended attribute area reading.
*/
int
ffs_ea_read(struct vnode *vp, struct uio *uio, int ioflag)
{
struct inode *ip;
struct ufs2_dinode *dp;
struct fs *fs;
struct buf *bp;
daddr64_t lbn, nextlbn;
off_t ealeft;
int error, size, xfersize;
size_t oresid;
ip = VTOI(vp);
fs = ip->i_fs;
dp = ip->i_din2;
error = 0;
oresid = uio->uio_resid;
ealeft = dp->di_extsize;
/*
* Loop over the amount of data requested by the caller, stopping only
* if an error occurs. By default, we always try to copy a file system
* block worth of bytes per iteration ('xfersize'). Check this value
* against what is left to be copied ('uio->uio_resid'), and the amount
* of bytes past our current position in the extended attribute area
* ('ealeft').
*/
while (uio->uio_resid > 0) {
ealeft -= uio->uio_offset;
if (ealeft <= 0)
break;
xfersize = fs->fs_bsize;
if (uio->uio_resid < xfersize)
xfersize = uio->uio_resid;
if (ealeft < xfersize)
xfersize = ealeft;
/*
* Get the corresponding logical block number. Read it in,
* doing read-ahead if possible.
*/
lbn = lblkno(fs, uio->uio_offset);
size = sblksize(fs, dp->di_extsize, lbn);
nextlbn = lbn + 1;
if (lblktosize(fs, nextlbn) >= dp->di_extsize)
error = bread(vp, -1 - lbn, size, NOCRED, &bp);
else {
int nextsize = sblksize(fs, dp->di_extsize, nextlbn);
nextlbn = -1 - nextlbn;
error = breadn(vp, -1 - lbn,
size, &nextlbn, &nextsize, 1, NOCRED, &bp);
}
if (error) {
brelse(bp);
break;
}
/* Check for short-reads. */
if (bp->b_resid) {
brelse(bp);
error = EIO;
break;
}
/* Finally, copy out the data, and release the buffer. */
error = uiomove(bp->b_data, xfersize, uio);
brelse(bp);
if (error)
break;
}
if ((error == 0 || uio->uio_resid != oresid) &&
(vp->v_mount->mnt_flag & MNT_NOATIME) == 0)
ip->i_flag |= IN_ACCESS;
return (error);
}
/*
* Vnode op for reading.
*/
int
cd9660_read(void *v)
{
struct vop_read_args *ap = v;
struct vnode *vp = ap->a_vp;
register struct uio *uio = ap->a_uio;
register struct iso_node *ip = VTOI(vp);
register struct iso_mnt *imp;
struct buf *bp;
daddr_t lbn, rablock;
off_t diff;
int error = 0;
long size, n, on;
if (uio->uio_resid == 0)
return (0);
if (uio->uio_offset < 0)
return (EINVAL);
ip->i_flag |= IN_ACCESS;
imp = ip->i_mnt;
do {
struct cluster_info *ci = &ip->i_ci;
lbn = lblkno(imp, uio->uio_offset);
on = blkoff(imp, uio->uio_offset);
n = min((u_int)(imp->logical_block_size - on),
uio->uio_resid);
diff = (off_t)ip->i_size - uio->uio_offset;
if (diff <= 0)
return (0);
if (diff < n)
n = diff;
size = blksize(imp, ip, lbn);
rablock = lbn + 1;
#define MAX_RA 32
if (ci->ci_lastr + 1 == lbn) {
struct ra {
daddr_t blks[MAX_RA];
int sizes[MAX_RA];
} *ra;
int i;
ra = malloc(sizeof *ra, M_TEMP, M_WAITOK);
for (i = 0; i < MAX_RA &&
lblktosize(imp, (rablock + i)) < ip->i_size;
i++) {
ra->blks[i] = rablock + i;
ra->sizes[i] = blksize(imp, ip, rablock + i);
}
error = breadn(vp, lbn, size, ra->blks,
ra->sizes, i, &bp);
free(ra, M_TEMP, 0);
} else
error = bread(vp, lbn, size, &bp);
ci->ci_lastr = lbn;
n = min(n, size - bp->b_resid);
if (error) {
brelse(bp);
return (error);
}
error = uiomovei(bp->b_data + on, (int)n, uio);
brelse(bp);
} while (error == 0 && uio->uio_resid > 0 && n != 0);
return (error);
}
static int journal_insertfile( ino_t ino ) {
struct ufs1_dinode *dp1;
struct ufs2_dinode *dp2;
void *ip;
ufs2_daddr_t nblk;
ufs2_daddr_t blk;
ufs_lbn_t lbn;
int size;
int mode;
int off;
if ( getino( &disk, &ip, ROOTINO, &mode ) != 0 ) {
warn( "Failed to get root inode" );
sbdirty();
return ( -1 );
}
dp2 = ip;
dp1 = ip;
blk = 0;
size = 0;
nblk = journal_balloc();
if ( nblk <= 0 ) return ( -1 );
if ( sblock.fs_magic == FS_UFS1_MAGIC ) {
lbn = lblkno( &sblock, dp1->di_size );
off = blkoff( &sblock, dp1->di_size );
blk = dp1->di_db[lbn];
size = sblksize( &sblock, (off_t) dp1->di_size, lbn );
} else {
lbn = lblkno( &sblock, dp2->di_size );
off = blkoff( &sblock, dp2->di_size );
blk = dp2->di_db[lbn];
size = sblksize( &sblock, (off_t) dp2->di_size, lbn );
}
if ( off != 0 ) {
if ( dir_extend( blk, nblk, off, ino ) == -1 ) return ( -1 );
} else {
blk = 0;
if ( dir_insert( nblk, 0, ino ) == -1 ) return ( -1 );
}
if ( sblock.fs_magic == FS_UFS1_MAGIC ) {
dp1->di_blocks += ( sblock.fs_bsize - size ) / DEV_BSIZE;
dp1->di_db[lbn] = nblk;
dp1->di_size = lblktosize( &sblock, lbn + 1 );
} else {
dp2->di_blocks += ( sblock.fs_bsize - size ) / DEV_BSIZE;
dp2->di_db[lbn] = nblk;
dp2->di_size = lblktosize( &sblock, lbn + 1 );
}
if ( putino( &disk ) < 0 ) {
warn( "Failed to write root inode" );
return ( -1 );
}
if ( cgwrite( &disk ) < 0 ) {
warn( "Failed to write updated cg" );
sbdirty();
return ( -1 );
}
if ( blk ) {
if ( cgbfree( &disk, blk, size ) < 0 ) {
warn( "Failed to write cg" );
return ( -1 );
}
}
return ( 0 );
}
/*
* this function handles traditional block mapping
*/
static int
ext2_ind_read(struct vop_read_args *ap)
{
struct vnode *vp;
struct inode *ip;
struct uio *uio;
FS *fs;
struct buf *bp;
daddr_t lbn, nextlbn;
off_t bytesinfile;
long size, xfersize, blkoffset;
int error, orig_resid, seqcount;
seqcount = ap->a_ioflag >> IO_SEQSHIFT;
u_short mode;
vp = ap->a_vp;
ip = VTOI(vp);
mode = ip->i_mode;
uio = ap->a_uio;
#ifdef DIAGNOSTIC
if (uio->uio_rw != UIO_READ)
panic("%s: mode", READ_S);
if (vp->v_type == VLNK) {
if ((int)ip->i_size < vp->v_mount->mnt_maxsymlinklen)
panic("%s: short symlink", READ_S);
} else if (vp->v_type != VREG && vp->v_type != VDIR)
panic("%s: type %d", READ_S, vp->v_type);
#endif
orig_resid = uio->uio_resid;
KASSERT(orig_resid >= 0, ("ext2_read: uio->uio_resid < 0"));
if (orig_resid == 0)
return (0);
KASSERT(uio->uio_offset >= 0, ("ext2_read: uio->uio_offset < 0"));
fs = ip->I_FS;
if (uio->uio_offset < ip->i_size && uio->uio_offset >= fs->e2fs_maxfilesize)
return (EOVERFLOW);
for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) {
if ((bytesinfile = ip->i_size - uio->uio_offset) <= 0)
break;
lbn = lblkno(fs, uio->uio_offset);
nextlbn = lbn + 1;
size = BLKSIZE(fs, ip, lbn);
blkoffset = blkoff(fs, uio->uio_offset);
xfersize = fs->e2fs_fsize - blkoffset;
if (uio->uio_resid < xfersize)
xfersize = uio->uio_resid;
if (bytesinfile < xfersize)
xfersize = bytesinfile;
if (lblktosize(fs, nextlbn) >= ip->i_size)
error = bread(vp, lbn, size, NOCRED, &bp);
else if ((vp->v_mount->mnt_flag & MNT_NOCLUSTERR) == 0)
error = cluster_read(vp, ip->i_size, lbn, size,
NOCRED, blkoffset + uio->uio_resid, seqcount, &bp);
else if (seqcount > 1) {
int nextsize = BLKSIZE(fs, ip, nextlbn);
error = breadn(vp, lbn,
size, &nextlbn, &nextsize, 1, NOCRED, &bp);
} else
error = bread(vp, lbn, size, NOCRED, &bp);
if (error) {
brelse(bp);
bp = NULL;
break;
}
/*
* We should only get non-zero b_resid when an I/O error
* has occurred, which should cause us to break above.
* However, if the short read did not cause an error,
* then we want to ensure that we do not uiomove bad
* or uninitialized data.
*/
size -= bp->b_resid;
if (size < xfersize) {
if (size == 0)
break;
xfersize = size;
}
error = uiomove((char *)bp->b_data + blkoffset,
(int)xfersize, uio);
if (error)
break;
bqrelse(bp);
}
if (bp != NULL)
bqrelse(bp);
if ((error == 0 || uio->uio_resid != orig_resid) &&
(vp->v_mount->mnt_flag & MNT_NOATIME) == 0)
ip->i_flag |= IN_ACCESS;
return (error);
}
/*
* Vnode op for reading.
*/
int
ffs_read(void *v)
{
struct vop_read_args *ap = v;
struct vnode *vp;
struct inode *ip;
struct uio *uio;
struct fs *fs;
struct buf *bp;
daddr_t lbn, nextlbn;
off_t bytesinfile;
long size, xfersize, blkoffset;
mode_t mode;
int error;
vp = ap->a_vp;
ip = VTOI(vp);
mode = DIP(ip, mode);
uio = ap->a_uio;
#ifdef DIAGNOSTIC
if (uio->uio_rw != UIO_READ)
panic("ffs_read: mode");
if (vp->v_type == VLNK) {
if ((int)DIP(ip, size) < vp->v_mount->mnt_maxsymlinklen ||
(vp->v_mount->mnt_maxsymlinklen == 0 &&
DIP(ip, blocks) == 0))
panic("ffs_read: short symlink");
} else if (vp->v_type != VREG && vp->v_type != VDIR)
panic("ffs_read: type %d", vp->v_type);
#endif
fs = ip->i_fs;
if ((u_int64_t)uio->uio_offset > fs->fs_maxfilesize)
return (EFBIG);
if (uio->uio_resid == 0)
return (0);
for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) {
if ((bytesinfile = DIP(ip, size) - uio->uio_offset) <= 0)
break;
lbn = lblkno(fs, uio->uio_offset);
nextlbn = lbn + 1;
size = fs->fs_bsize; /* WAS blksize(fs, ip, lbn); */
blkoffset = blkoff(fs, uio->uio_offset);
xfersize = fs->fs_bsize - blkoffset;
if (uio->uio_resid < xfersize)
xfersize = uio->uio_resid;
if (bytesinfile < xfersize)
xfersize = bytesinfile;
if (lblktosize(fs, nextlbn) >= DIP(ip, size))
error = bread(vp, lbn, size, &bp);
else if (lbn - 1 == ip->i_ci.ci_lastr) {
error = bread_cluster(vp, lbn, size, &bp);
} else
error = bread(vp, lbn, size, &bp);
if (error)
break;
ip->i_ci.ci_lastr = lbn;
/*
* We should only get non-zero b_resid when an I/O error
* has occurred, which should cause us to break above.
* However, if the short read did not cause an error,
* then we want to ensure that we do not uiomove bad
* or uninitialized data.
*/
size -= bp->b_resid;
if (size < xfersize) {
if (size == 0)
break;
xfersize = size;
}
error = uiomovei(bp->b_data + blkoffset, (int)xfersize, uio);
if (error)
break;
brelse(bp);
}
if (bp != NULL)
brelse(bp);
if (!(vp->v_mount->mnt_flag & MNT_NOATIME) ||
(ip->i_flag & (IN_CHANGE | IN_UPDATE))) {
ip->i_flag |= IN_ACCESS;
}
return (error);
}
/*
* 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;
}
/*
* Vnode op for reading.
*/
static int
ext2_read(struct vop_read_args *ap)
{
struct vnode *vp;
struct inode *ip;
struct uio *uio;
struct m_ext2fs *fs;
struct buf *bp;
daddr_t lbn, nextlbn;
off_t bytesinfile;
long size, xfersize, blkoffset;
int error, orig_resid, seqcount;
int ioflag;
vp = ap->a_vp;
uio = ap->a_uio;
ioflag = ap->a_ioflag;
seqcount = ap->a_ioflag >> IO_SEQSHIFT;
ip = VTOI(vp);
#ifdef INVARIANTS
if (uio->uio_rw != UIO_READ)
panic("%s: mode", "ext2_read");
if (vp->v_type == VLNK) {
if ((int)ip->i_size < vp->v_mount->mnt_maxsymlinklen)
panic("%s: short symlink", "ext2_read");
} else if (vp->v_type != VREG && vp->v_type != VDIR)
panic("%s: type %d", "ext2_read", vp->v_type);
#endif
orig_resid = uio->uio_resid;
KASSERT(orig_resid >= 0, ("ext2_read: uio->uio_resid < 0"));
if (orig_resid == 0)
return (0);
KASSERT(uio->uio_offset >= 0, ("ext2_read: uio->uio_offset < 0"));
fs = ip->i_e2fs;
if (uio->uio_offset < ip->i_size &&
uio->uio_offset >= fs->e2fs_maxfilesize)
return (EOVERFLOW);
for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) {
if ((bytesinfile = ip->i_size - uio->uio_offset) <= 0)
break;
lbn = lblkno(fs, uio->uio_offset);
nextlbn = lbn + 1;
size = blksize(fs, ip, lbn);
blkoffset = blkoff(fs, uio->uio_offset);
xfersize = fs->e2fs_fsize - blkoffset;
if (uio->uio_resid < xfersize)
xfersize = uio->uio_resid;
if (bytesinfile < xfersize)
xfersize = bytesinfile;
if (lblktosize(fs, nextlbn) >= ip->i_size)
error = bread(vp, lbn, size, NOCRED, &bp);
else if ((vp->v_mount->mnt_flag & MNT_NOCLUSTERR) == 0) {
error = cluster_read(vp, ip->i_size, lbn, size,
NOCRED, blkoffset + uio->uio_resid, seqcount,
0, &bp);
} else if (seqcount > 1) {
int nextsize = blksize(fs, ip, nextlbn);
error = breadn(vp, lbn,
size, &nextlbn, &nextsize, 1, NOCRED, &bp);
} else
error = bread(vp, lbn, size, NOCRED, &bp);
if (error) {
brelse(bp);
bp = NULL;
break;
}
/*
* If IO_DIRECT then set B_DIRECT for the buffer. This
* will cause us to attempt to release the buffer later on
* and will cause the buffer cache to attempt to free the
* underlying pages.
*/
if (ioflag & IO_DIRECT)
bp->b_flags |= B_DIRECT;
/*
* We should only get non-zero b_resid when an I/O error
* has occurred, which should cause us to break above.
* However, if the short read did not cause an error,
* then we want to ensure that we do not uiomove bad
* or uninitialized data.
*/
size -= bp->b_resid;
if (size < xfersize) {
if (size == 0)
break;
xfersize = size;
}
error = uiomove((char *)bp->b_data + blkoffset,
(int)xfersize, uio);
if (error)
break;
if (ioflag & (IO_VMIO|IO_DIRECT)) {
/*
* If it's VMIO or direct I/O, then we don't
* need the buf, mark it available for
* freeing. If it's non-direct VMIO, the VM has
* the data.
*/
bp->b_flags |= B_RELBUF;
brelse(bp);
} else {
/*
* Otherwise let whoever
* made the request take care of
* freeing it. We just queue
* it onto another list.
*/
bqrelse(bp);
}
}
/*
* This can only happen in the case of an error
* because the loop above resets bp to NULL on each iteration
* and on normal completion has not set a new value into it.
* so it must have come from a 'break' statement
*/
if (bp != NULL) {
if (ioflag & (IO_VMIO|IO_DIRECT)) {
bp->b_flags |= B_RELBUF;
brelse(bp);
} else {
bqrelse(bp);
}
}
if ((error == 0 || uio->uio_resid != orig_resid) &&
(vp->v_mount->mnt_flag & MNT_NOATIME) == 0)
ip->i_flag |= IN_ACCESS;
return (error);
}
int
ffs2_balloc(struct inode *ip, off_t off, int size, struct ucred *cred,
int flags, struct buf **bpp)
{
daddr_t lbn, lastlbn, nb, newb, *blkp;
daddr_t pref, *allocblk, allociblk[NIADDR + 1];
daddr_t *bap, *allocib;
int deallocated, osize, nsize, num, i, error, unwindidx, r;
struct buf *bp, *nbp;
struct indir indirs[NIADDR + 2];
struct fs *fs;
struct vnode *vp;
struct proc *p;
vp = ITOV(ip);
fs = ip->i_fs;
p = curproc;
unwindidx = -1;
lbn = lblkno(fs, off);
size = blkoff(fs, off) + size;
if (size > fs->fs_bsize)
panic("ffs2_balloc: block too big");
if (bpp != NULL)
*bpp = NULL;
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_ffs2_size);
if (lastlbn < NDADDR && lastlbn < lbn) {
nb = lastlbn;
osize = blksize(fs, ip, nb);
if (osize < fs->fs_bsize && osize > 0) {
error = ffs_realloccg(ip, nb, ffs2_blkpref(ip,
lastlbn, nb, &ip->i_ffs2_db[0]), osize,
(int) fs->fs_bsize, cred, bpp, &newb);
if (error)
return (error);
if (DOINGSOFTDEP(vp))
softdep_setup_allocdirect(ip, nb, newb,
ip->i_ffs2_db[nb], fs->fs_bsize, osize,
bpp ? *bpp : NULL);
ip->i_ffs2_size = lblktosize(fs, nb + 1);
uvm_vnp_setsize(vp, ip->i_ffs2_size);
ip->i_ffs2_db[nb] = newb;
ip->i_flag |= IN_CHANGE | IN_UPDATE;
if (bpp) {
if (flags & B_SYNC)
bwrite(*bpp);
else
bawrite(*bpp);
}
}
}
/*
* The first NDADDR blocks are direct.
*/
if (lbn < NDADDR) {
nb = ip->i_ffs2_db[lbn];
if (nb != 0 && ip->i_ffs2_size >= lblktosize(fs, lbn + 1)) {
/*
* The direct block is already allocated and the file
* extends past this block, thus this must be a whole
* block. Just read it, if requested.
*/
if (bpp != NULL) {
error = bread(vp, lbn, fs->fs_bsize, bpp);
if (error) {
brelse(*bpp);
return (error);
}
}
return (0);
}
if (nb != 0) {
/*
* Consider the need to allocate a fragment.
*/
osize = fragroundup(fs, blkoff(fs, ip->i_ffs2_size));
nsize = fragroundup(fs, size);
if (nsize <= osize) {
/*
* The existing block is already at least as
* big as we want. Just read it, if requested.
*/
if (bpp != NULL) {
error = bread(vp, lbn, fs->fs_bsize,
bpp);
if (error) {
brelse(*bpp);
return (error);
}
(*bpp)->b_bcount = osize;
}
return (0);
} else {
/*
* The existing block is smaller than we want,
* grow it.
*/
error = ffs_realloccg(ip, lbn,
ffs2_blkpref(ip, lbn, (int) lbn,
&ip->i_ffs2_db[0]), osize, nsize, cred,
bpp, &newb);
if (error)
return (error);
if (DOINGSOFTDEP(vp))
softdep_setup_allocdirect(ip, lbn,
newb, nb, nsize, osize,
bpp ? *bpp : NULL);
}
} else {
/*
* The block was not previously allocated, allocate a
* new block or fragment.
*/
if (ip->i_ffs2_size < lblktosize(fs, lbn + 1))
nsize = fragroundup(fs, size);
else
nsize = fs->fs_bsize;
error = ffs_alloc(ip, lbn, ffs2_blkpref(ip, lbn,
(int) lbn, &ip->i_ffs2_db[0]), nsize, cred, &newb);
if (error)
return (error);
if (bpp != NULL) {
bp = getblk(vp, lbn, fs->fs_bsize, 0, 0);
if (nsize < fs->fs_bsize)
bp->b_bcount = nsize;
bp->b_blkno = fsbtodb(fs, newb);
if (flags & B_CLRBUF)
clrbuf(bp);
*bpp = bp;
}
if (DOINGSOFTDEP(vp))
softdep_setup_allocdirect(ip, lbn, newb, 0,
nsize, 0, bpp ? *bpp : NULL);
}
ip->i_ffs2_db[lbn] = newb;
ip->i_flag |= IN_CHANGE | IN_UPDATE;
return (0);
}
/*
* Determine the number of levels of indirection.
*/
pref = 0;
error = ufs_getlbns(vp, lbn, indirs, &num);
if (error)
return (error);
#ifdef DIAGNOSTIC
if (num < 1)
panic("ffs2_balloc: ufs_bmaparray returned indirect block");
#endif
/*
* Fetch the first indirect block allocating it necessary.
*/
--num;
nb = ip->i_ffs2_ib[indirs[0].in_off];
allocib = NULL;
allocblk = allociblk;
if (nb == 0) {
pref = ffs2_blkpref(ip, lbn, -indirs[0].in_off - 1, NULL);
error = ffs_alloc(ip, lbn, pref, (int) fs->fs_bsize, cred,
&newb);
if (error)
goto fail;
nb = newb;
*allocblk++ = nb;
bp = getblk(vp, indirs[1].in_lbn, fs->fs_bsize, 0, 0);
bp->b_blkno = fsbtodb(fs, nb);
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.
*/
error = bwrite(bp);
if (error)
goto fail;
}
unwindidx = 0;
allocib = &ip->i_ffs2_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, indirs[i].in_lbn, (int)fs->fs_bsize, &bp);
if (error) {
brelse(bp);
goto fail;
}
bap = (int64_t *) bp->b_data;
nb = bap[indirs[i].in_off];
if (i == num)
break;
i++;
if (nb != 0) {
brelse(bp);
continue;
}
if (pref == 0)
pref = ffs2_blkpref(ip, lbn, i - num - 1, NULL);
error = ffs_alloc(ip, lbn, pref, (int) fs->fs_bsize, cred,
&newb);
if (error) {
brelse(bp);
goto fail;
}
nb = newb;
*allocblk++ = nb;
nbp = getblk(vp, indirs[i].in_lbn, fs->fs_bsize, 0, 0);
nbp->b_blkno = fsbtodb(fs, nb);
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.
*/
error = bwrite(nbp);
if (error) {
brelse(bp);
goto fail;
}
}
if (unwindidx < 0)
unwindidx = i - 1;
bap[indirs[i - 1].in_off] = nb;
/*
* If required, write synchronously, otherwise use delayed
* write.
*/
if (flags & B_SYNC)
bwrite(bp);
else
bdwrite(bp);
}
/*
* Get the data block, allocating if necessary.
*/
if (nb == 0) {
pref = ffs2_blkpref(ip, lbn, indirs[num].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;
if (bpp != NULL) {
nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0);
nbp->b_blkno = fsbtodb(fs, nb);
if (flags & B_CLRBUF)
clrbuf(nbp);
*bpp = nbp;
}
if (DOINGSOFTDEP(vp))
softdep_setup_allocindir_page(ip, lbn, bp,
indirs[num].in_off, nb, 0, bpp ? *bpp : NULL);
bap[indirs[num].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
bdwrite(bp);
return (0);
}
brelse(bp);
if (bpp != NULL) {
if (flags & B_CLRBUF) {
error = bread(vp, lbn, (int)fs->fs_bsize, &nbp);
if (error) {
brelse(nbp);
goto fail;
}
} else {
nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0);
nbp->b_blkno = fsbtodb(fs, nb);
clrbuf(nbp);
}
*bpp = nbp;
}
return (0);
fail:
/*
* If we have failed to allocate any blocks, simply return the error.
* This is the usual case and avoids the need to fsync the file.
*/
if (allocblk == allociblk && allocib == NULL && unwindidx == -1)
return (error);
/*
* If we have failed part way through block allocation, we have to
* deallocate any indirect blocks that we have allocated. We have to
* fsync the file before we start to get rid of all of its
* dependencies so that we do not leave them dangling. We have to sync
* it at the end so that the softdep code does not find any untracked
* changes. Although this is really slow, running out of disk space is
* not expected to be a common occurrence. The error return from fsync
* is ignored as we already have an error to return to the user.
*/
VOP_FSYNC(vp, p->p_ucred, MNT_WAIT, p);
if (unwindidx >= 0) {
/*
* First write out any buffers we've created to resolve their
* softdeps. This must be done in reverse order of creation so
* that we resolve the dependencies in one pass.
* Write the cylinder group buffers for these buffers too.
*/
for (i = num; i >= unwindidx; i--) {
if (i == 0)
break;
bp = getblk(vp, indirs[i].in_lbn, (int) fs->fs_bsize,
0, 0);
if (bp->b_flags & B_DELWRI) {
nb = fsbtodb(fs, cgtod(fs, dtog(fs,
dbtofsb(fs, bp->b_blkno))));
bwrite(bp);
bp = getblk(ip->i_devvp, nb,
(int) fs->fs_cgsize, 0, 0);
if (bp->b_flags & B_DELWRI)
bwrite(bp);
else {
bp->b_flags |= B_INVAL;
brelse(bp);
}
} else {
bp->b_flags |= B_INVAL;
brelse(bp);
}
}
if (DOINGSOFTDEP(vp) && unwindidx == 0) {
ip->i_flag |= IN_CHANGE | IN_UPDATE;
ffs_update(ip, 1);
}
/*
* Now that any dependencies that we created have been
* resolved, we can undo the partial allocation.
*/
if (unwindidx == 0) {
*allocib = 0;
ip->i_flag |= IN_CHANGE | IN_UPDATE;
if (DOINGSOFTDEP(vp))
ffs_update(ip, 1);
} else {
r = bread(vp, indirs[unwindidx].in_lbn,
(int)fs->fs_bsize, &bp);
if (r)
panic("ffs2_balloc: unwind failed");
bap = (int64_t *) bp->b_data;
bap[indirs[unwindidx].in_off] = 0;
bwrite(bp);
}
for (i = unwindidx + 1; i <= num; i++) {
bp = getblk(vp, indirs[i].in_lbn, (int)fs->fs_bsize, 0,
0);
bp->b_flags |= B_INVAL;
brelse(bp);
}
}
for (deallocated = 0, blkp = allociblk; blkp < allocblk; blkp++) {
ffs_blkfree(ip, *blkp, fs->fs_bsize);
deallocated += fs->fs_bsize;
}
if (deallocated) {
/*
* Restore user's disk quota because allocation failed.
*/
(void) ufs_quota_free_blocks(ip, btodb(deallocated), cred);
ip->i_ffs2_blocks -= btodb(deallocated);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
}
VOP_FSYNC(vp, p->p_ucred, MNT_WAIT, p);
return (error);
}
