block-vhd.c

xen source 推出最新的VHD操作工具VHD-UTIL 实现源码,超强

	allocated = 0;
	full      = 0;

	for (i = 0; i < s->bat.bat.entries; i++) {
		if (bat_entry(s, i) != DD_BLK_UNUSED)
			allocated++;
		if (test_batmap(s, i))
			full++;
	}

	DPRINTF("%s version: %s 0x%08x, b: %u, a: %u, f: %u, n: %llu\n",
		s->vhd.file, buf, s->vhd.footer.crtr_ver, s->bat.bat.entries,
		allocated, full, s->next_db);
}

static int
__vhd_open(td_driver_t *driver, const char *name, vhd_flag_t flags)
{
        int i, o_flags, err;
	struct vhd_state *s;

        DBG(TLOG_INFO, "vhd_open: %s\n", name);
	if (test_vhd_flag(flags, VHD_FLAG_OPEN_STRICT))
		libvhd_set_log_level(1);

	s = (struct vhd_state *)driver->data;
	memset(s, 0, sizeof(struct vhd_state));

	s->flags  = flags;
	s->driver = driver;

	err = vhd_initialize(s);
	if (err)
		return err;

	o_flags = ((test_vhd_flag(flags, VHD_FLAG_OPEN_RDONLY)) ? 
		   VHD_OPEN_RDONLY : VHD_OPEN_RDWR);

	err = vhd_open(&s->vhd, name, o_flags);
	if (err) {
		libvhd_set_log_level(1);
		err = vhd_open(&s->vhd, name, o_flags);
		if (err) {
			EPRINTF("Unable to open [%s] (%d)!\n", name, err);
			return err;
		}
	}

	err = vhd_check_version(s);
	if (err)
		goto fail;

	s->spb = s->spp = 1;

	if (vhd_type_dynamic(&s->vhd)) {
		err = vhd_initialize_dynamic_disk(s);
		if (err)
			goto fail;
	}

	vhd_log_open(s);

	SPB = s->spb;

	s->vreq_free_count = VHD_REQS_DATA;
	for (i = 0; i < VHD_REQS_DATA; i++)
		s->vreq_free[i] = s->vreq_list + i;

	driver->info.size        = s->vhd.footer.curr_size >> VHD_SECTOR_SHIFT;
	driver->info.sector_size = VHD_SECTOR_SIZE;
	driver->info.info        = 0;

        DBG(TLOG_INFO, "vhd_open: done (sz:%llu, sct:%lu, inf:%u)\n",
	    driver->info.size, driver->info.sector_size, driver->info.info);

	if (test_vhd_flag(flags, VHD_FLAG_OPEN_STRICT) && 
	    !test_vhd_flag(flags, VHD_FLAG_OPEN_RDONLY)) {
		err = vhd_kill_footer(s);
		if (err) {
			DPRINTF("ERROR killing footer: %d\n", err);
			goto fail;
		}
		s->writes++;
	}

        return 0;

 fail:
	vhd_free_bat(s);
	vhd_free_bitmap_cache(s);
	vhd_close(&s->vhd);
	vhd_free(s);
	return err;
}

static int
_vhd_open(td_driver_t *driver, const char *name, td_flag_t flags)
{
	vhd_flag_t vhd_flags = 0;

	if (flags & TD_OPEN_RDONLY)
		vhd_flags |= VHD_FLAG_OPEN_RDONLY;
	if (flags & TD_OPEN_QUIET)
		vhd_flags |= VHD_FLAG_OPEN_QUIET;
	if (flags & TD_OPEN_STRICT)
		vhd_flags |= VHD_FLAG_OPEN_STRICT;
	if (flags & TD_OPEN_QUERY)
		vhd_flags |= (VHD_FLAG_OPEN_QUERY  |
			      VHD_FLAG_OPEN_QUIET  |
			      VHD_FLAG_OPEN_RDONLY |
			      VHD_FLAG_OPEN_NO_CACHE);

	/* pre-allocate for all but NFS and LVM storage */
	if (driver->storage != TAPDISK_STORAGE_TYPE_NFS &&
	    driver->storage != TAPDISK_STORAGE_TYPE_LVM)
		vhd_flags |= VHD_FLAG_OPEN_PREALLOCATE;

	return __vhd_open(driver, name, vhd_flags);
}

static void
vhd_log_close(struct vhd_state *s)
{
	uint32_t i, allocated, full;

	if (test_vhd_flag(s->flags, VHD_FLAG_OPEN_QUIET))
		return;

	allocated = 0;
	full      = 0;

	for (i = 0; i < s->bat.bat.entries; i++) {
		if (bat_entry(s, i) != DD_BLK_UNUSED)
			allocated++;
		if (test_batmap(s, i))
			full++;
	}

	DPRINTF("%s: b: %u, a: %u, f: %u, n: %llu\n",
		s->vhd.file, s->bat.bat.entries, allocated, full, s->next_db);
}

static int
_vhd_close(td_driver_t *driver)
{
	int err;
	struct vhd_state *s;
	struct vhd_bitmap *bm;
	
	DBG(TLOG_WARN, "vhd_close\n");
	s = (struct vhd_state *)driver->data;

	/* don't write footer if tapdisk is read-only */
	if (test_vhd_flag(s->flags, VHD_FLAG_OPEN_RDONLY))
		goto free;
	
	/* 
	 * write footer if:
	 *   - we killed it on open (opened with strict) 
	 *   - we've written data since opening
	 */
	if (test_vhd_flag(s->flags, VHD_FLAG_OPEN_STRICT) || s->writes) {
		memcpy(&s->vhd.bat, &s->bat.bat, sizeof(vhd_bat_t));
		err = vhd_write_footer(&s->vhd, &s->vhd.footer);
		memset(&s->vhd.bat, 0, sizeof(vhd_bat_t));

		if (err)
			EPRINTF("writing %s footer: %d\n", s->vhd.file, err);

		if (!vhd_has_batmap(&s->vhd))
			goto free;

		err = vhd_write_batmap(&s->vhd, &s->bat.batmap);
		if (err)
			EPRINTF("writing %s batmap: %d\n", s->vhd.file, err);
	}

 free:
	vhd_log_close(s);
	vhd_free_bat(s);
	vhd_free_bitmap_cache(s);
	vhd_close(&s->vhd);
	vhd_free(s);

	memset(s, 0, sizeof(struct vhd_state));

	return 0;
}

int
vhd_validate_parent(td_driver_t *child_driver,
		    td_driver_t *parent_driver, td_flag_t flags)
{
	struct stat stats;
	struct vhd_state *child  = (struct vhd_state *)child_driver->data;
	struct vhd_state *parent;

	if (parent_driver->type != DISK_TYPE_VHD) {
		if (child_driver->type != DISK_TYPE_VHD)
			return -EINVAL;
		if (child->vhd.footer.type != HD_TYPE_DIFF)
			return -EINVAL;
		if (!vhd_parent_raw(&child->vhd))
			return -EINVAL;
		return 0;
	}

	parent = (struct vhd_state *)parent_driver->data;

	/* 
	 * This check removed because of cases like:
	 *   - parent VHD marked as 'hidden'
	 *   - parent VHD modified during coalesce
	 */
	/*
	if (stat(parent->vhd.file, &stats)) {
		DPRINTF("ERROR stating parent file %s\n", parent->vhd.file);
		return -errno;
	}

	if (child->hdr.prt_ts != vhd_time(stats.st_mtime)) {
		DPRINTF("ERROR: parent file has been modified since "
			"snapshot.  Child image no longer valid.\n");
		return -EINVAL;
	}
	*/

	if (uuid_compare(child->vhd.header.prt_uuid, parent->vhd.footer.uuid)) {
		DPRINTF("ERROR: %s: %s, %s: parent uuid has changed since "
			"snapshot.  Child image no longer valid.\n",
			__func__, child->vhd.file, parent->vhd.file);
		return -EINVAL;
	}

	/* TODO: compare sizes */
	
	return 0;
}

int
vhd_get_parent_id(td_driver_t *driver, td_disk_id_t *id)
{
	int err;
	char *parent;
	struct vhd_state *s;

	DBG(TLOG_DBG, "\n");
	memset(id, 0, sizeof(td_disk_id_t));

	s = (struct vhd_state *)driver->data;

	if (s->vhd.footer.type != HD_TYPE_DIFF)
		return TD_NO_PARENT;

	err = vhd_parent_locator_get(&s->vhd, &parent);
	if (err)
		return err;

	id->name       = parent;
	id->drivertype = DISK_TYPE_VHD;
	if (vhd_parent_raw(&s->vhd)) {
		DPRINTF("VHD: parent is raw\n");
		id->drivertype = DISK_TYPE_AIO;
	}
	return 0;
}

static inline void
clear_req_list(struct vhd_req_list *list)
{
	list->head = list->tail = NULL;
}

static inline void
add_to_tail(struct vhd_req_list *list, struct vhd_request *e)
{
	if (!list->head) 
		list->head = list->tail = e;
	else 
		list->tail = list->tail->next = e;
}

static inline int
remove_from_req_list(struct vhd_req_list *list, struct vhd_request *e)
{
	struct vhd_request *i = list->head;

	if (list->head == e) {
		if (list->tail == e)
			clear_req_list(list);
		else
			list->head = list->head->next;
		return 0;
	}

	while (i->next) {
		if (i->next == e) {
			if (list->tail == e) {
				i->next = NULL;
				list->tail = i;
			} else
				i->next = i->next->next;
			return 0;
		}
		i = i->next;
	}

	return -EINVAL;
}

static inline void
init_vhd_request(struct vhd_state *s, struct vhd_request *req)
{
	memset(req, 0, sizeof(struct vhd_request));
	req->state = s;
}

static inline void
init_tx(struct vhd_transaction *tx)
{
	memset(tx, 0, sizeof(struct vhd_transaction));
}

static inline void
add_to_transaction(struct vhd_transaction *tx, struct vhd_request *r)
{
	ASSERT(!tx->closed);

	r->tx = tx;
	tx->started++;
	add_to_tail(&tx->requests, r);
	set_vhd_flag(tx->status, VHD_FLAG_TX_LIVE);

	DBG(TLOG_DBG, "blk: 0x%04"PRIx64", lsec: 0x%08"PRIx64", tx: %p, "
	    "started: %d, finished: %d, status: %u\n",
	    r->treq.sec / SPB, r->treq.sec, tx,
	    tx->started, tx->finished, tx->status);
}

static inline int
transaction_completed(struct vhd_transaction *tx)
{
	return (tx->started == tx->finished);
}

static inline void
init_bat(struct vhd_state *s)
{
	s->bat.req.tx     = NULL;
	s->bat.req.next   = NULL;
	s->bat.req.error  = 0;
	s->bat.pbw_blk    = 0;
	s->bat.pbw_offset = 0;
	s->bat.status     = 0;
}

static inline void
lock_bat(struct vhd_state *s)
{
	set_vhd_flag(s->bat.status, VHD_FLAG_BAT_LOCKED);
}

static inline void
unlock_bat(struct vhd_state *s)
{
	clear_vhd_flag(s->bat.status, VHD_FLAG_BAT_LOCKED);
}

static inline int
bat_locked(struct vhd_state *s)
{
	return test_vhd_flag(s->bat.status, VHD_FLAG_BAT_LOCKED);
}

static inline void
init_vhd_bitmap(struct vhd_state *s, struct vhd_bitmap *bm)
{
	bm->blk    = 0;
	bm->seqno  = 0;
	bm->status = 0;
	init_tx(&bm->tx);
	clear_req_list(&bm->queue);
	clear_req_list(&bm->waiting);
	memset(bm->map, 0, vhd_sectors_to_bytes(s->bm_secs));
	memset(bm->shadow, 0, vhd_sectors_to_bytes(s->bm_secs));
	init_vhd_request(s, &bm->req);
}

static inline struct vhd_bitmap *
get_bitmap(struct vhd_state *s, uint32_t block)
{
	int i;
	struct vhd_bitmap *bm;

	for (i = 0; i < VHD_CACHE_SIZE; i++) {
		bm = s->bitmap[i];
		if (bm && bm->blk == block)
			return bm;
	}

	return NULL;
}

static inline void
lock_bitmap(struct vhd_bitmap *bm)
{
	set_vhd_flag(bm->status, VHD_FLAG_BM_LOCKED);
}

static inline void
unlock_bitmap(struct vhd_bitmap *bm)
{
	clear_vhd_flag(bm->status, VHD_FLAG_BM_LOCKED);
}

static inline int
bitmap_locked(struct vhd_bitmap *bm)
{
	return test_vhd_flag(bm->status, VHD_FLAG_BM_LOCKED);
}

static inline int
bitmap_valid(struct vhd_bitmap *bm)
{
	return !test_vhd_flag(bm->status, VHD_FLAG_BM_READ_PENDING);
}

static inline int
bitmap_in_use(struct vhd_bitmap *bm)
{
	return (test_vhd_flag(bm->status, VHD_FLAG_BM_READ_PENDING)  ||
		test_vhd_flag(bm->status, VHD_FLAG_BM_WRITE_PENDING) ||
		test_vhd_flag(bm->tx.status, VHD_FLAG_TX_UPDATE_BAT) ||
		bm->waiting.head || bm->tx.requests.head || bm->queue.head);
}

static inline int
bitmap_full(struct vhd_state *s, struct vhd_bitmap *bm)
{
	int i, n;

	n = s->spb >> 3;
	for (i = 0; i < n; i++)
		if (bm->map[i] != (char)0xFF)
			return 0;

	DBG(TLOG_DBG, "bitmap 0x%04x full\n", bm->blk);
	return 1;
}

static struct vhd_bitmap *
remove_lru_bitmap(struct vhd_state *s)
{
	int i, idx = 0;
	u64 seq = s->bm_lru;
	struct vhd_bitmap *bm, *lru = NULL;

	for (i = 0; i < VHD_CACHE_SIZE; i++) {
		bm = s->bitmap[i];
		if (bm && bm->seqno < seq && !bitmap_locked(bm)) {
			idx = i;
			lru = bm;
			seq = lru->seqno;
		}
	}

	if (lru) {
		s->bitmap[idx] = NULL;
		ASSERT(!bitmap_in_use(lru));
	}

	return  lru;
}

static int
alloc_vhd_bitmap(struct vhd_state *s, struct vhd_bitmap **bitmap, uint32_t blk)
{
	struct vhd_bitmap *bm;
	
	*bitmap = NULL;

	if (s->bm_free_count > 0) {
		bm = s->bitmap_free[--s->bm_free_count];
	} else {
		bm = remove_lru_bitmap(s);
		if (!bm)
			return -EBUSY;
	}

	init_vhd_bitmap(s, bm);
	bm->blk = blk;
	*bitmap = bm;

	return 0;
}

static inline uint64_t
__bitmap_lru_seqno(struct vhd_state *s)
{
	int i;
	struct vhd_bitmap *bm;

	if (s->bm_lru == 0xffffffff) {
		s->bm_lru = 0;
		for (i = 0; i < VHD_CACHE_SIZE; i++) {
			bm = s->bitmap[i];
			if (bm) {
				bm->seqno >>= 1;
				if (bm->seqno > s->bm_lru)
					s->bm_lru = bm->seqno;
			}
		}
	}

	return ++s->bm_lru;
}

static inline void
touch_bitmap(struct vhd_state *s, struct vhd_bitmap *bm)
{
	bm->seqno = __bitmap_lru_seqno(s);
}

static inline void
install_bitmap(struct vhd_state *s, struct vhd_bitmap *bm)
{
	int i;
	for (i = 0; i < VHD_CACHE_SIZE; i++) {
		if (!s->bitmap[i]) {
			touch_bitmap(s, bm);
			s->bitmap[i] = bm;
			return;
		}
	}

	ASSERT(0);
}

static inline void
free_vhd_bitmap(struct vhd_state *s, struct vhd_bitmap *bm)
{
	int i;

	for (i = 0; i < VHD_CACHE_SIZE; i++)
		if (s->bitmap[i] == bm)
			break;

	ASSERT(!bitmap_locked(bm));
	ASSERT(!bitmap_in_use(bm));
	ASSERT(i < VHD_CACHE_SIZE);

	s->bitmap[i] = NULL;
	s->bitmap_free[s->bm_free_count++] = bm;
}

static int
read_bitmap_cache(struct vhd_state *s, uint64_t sector, uint8_t op)
{
	u32 blk, sec;
	struct vhd_bitmap *bm;

	/* in fixed disks, every block is present */
	if (s->vhd.footer.type == HD_TYPE_FIXED) 
		return VHD_BM_BIT_SET;

	blk = sector / s->spb;
	sec = sector % s->spb;

	if (blk > s->vhd.header.max_bat_size) {
		DPRINTF("ERROR: sec %"PRIu64" out of range, op = %d\n",
			sector, op);
		return -EINVAL;
	}

	if (bat_entry(s, blk) == DD_BLK_UNUSED) {
		if (op == VHD_OP_DATA_WRITE &&
		    s->bat.pbw_blk != blk && bat_locked(s))
			return VHD_BM_BAT_LOCKED;

		return VHD_BM_BAT_CLEAR;
	}