[PATCH 2/2] ocfs2: Implement group add for online resize

This patch adds the ability for a userspace program to request that a
properly formatted cluster group be added to the main allocation bitmap for
an Ocfs2 file system. The request is made via an ioctl, OCFS2_IOC_GROUP_ADD.
On a high level, this is similar to ext3, but we use a different ioctl as
the structure which has to be passed through is different.

During an online resize, tunefs.ocfs2 will format any new cluster groups
which must be added to complete the resize, and call OCFS2_IOC_GROUP_ADD on
each one. Kernel verifies that the core cluster group information is valid
and then does the work of linking it into the global allocation bitmap.

Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>

fs/ocfs2/ioctl.c

@@ -119,6 +119,7 @@ int ocfs2_ioctl(struct inode * inode, struct file * filp,
 	int new_clusters;
 	int status;
 	struct ocfs2_space_resv sr;
+	struct ocfs2_new_group_input input;
 
 	switch (cmd) {
 	case OCFS2_IOC_GETFLAGS:
@@ -147,6 +148,12 @@ int ocfs2_ioctl(struct inode * inode, struct file * filp,
 			return -EFAULT;
 
 		return ocfs2_group_extend(inode, new_clusters);
+	case OCFS2_IOC_GROUP_ADD:
+	case OCFS2_IOC_GROUP_ADD64:
+		if (copy_from_user(&input, (int __user *) arg, sizeof(input)))
+			return -EFAULT;
+
+		return ocfs2_group_add(inode, &input);
 	default:
 		return -ENOTTY;
 	}
@@ -170,6 +177,8 @@ long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg)
 	case OCFS2_IOC_UNRESVSP:
 	case OCFS2_IOC_UNRESVSP64:
 	case OCFS2_IOC_GROUP_EXTEND:
+	case OCFS2_IOC_GROUP_ADD:
+	case OCFS2_IOC_GROUP_ADD64:
 		break;
 	default:
 		return -ENOIOCTLCMD;

fs/ocfs2/journal.h

@@ -281,6 +281,9 @@ int                  ocfs2_journal_dirty_data(handle_t *handle,
 /* group extend. inode update and last group update. */
 #define OCFS2_GROUP_EXTEND_CREDITS	(OCFS2_INODE_UPDATE_CREDITS + 1)
 
+/* group add. inode update and the new group update. */
+#define OCFS2_GROUP_ADD_CREDITS	(OCFS2_INODE_UPDATE_CREDITS + 1)
+
 /* get one bit out of a suballocator: dinode + group descriptor +
  * prev. group desc. if we relink. */
 #define OCFS2_SUBALLOC_ALLOC (3)

fs/ocfs2/ocfs2_fs.h

@@ -231,7 +231,19 @@ struct ocfs2_space_resv {
 #define OCFS2_IOC_RESVSP64	_IOW ('X', 42, struct ocfs2_space_resv)
 #define OCFS2_IOC_UNRESVSP64	_IOW ('X', 43, struct ocfs2_space_resv)
 
+/* Used to pass group descriptor data when online resize is done */
+struct ocfs2_new_group_input {
+	__u64 group;		/* Group descriptor's blkno. */
+	__u32 clusters;		/* Total number of clusters in this group */
+	__u32 frees;		/* Total free clusters in this group */
+	__u16 chain;		/* Chain for this group */
+	__u16 reserved1;
+	__u32 reserved2;
+};
+
 #define OCFS2_IOC_GROUP_EXTEND	_IOW('o', 1, int)
+#define OCFS2_IOC_GROUP_ADD	_IOW('o', 2,struct ocfs2_new_group_input)
+#define OCFS2_IOC_GROUP_ADD64	_IOW('o', 3,struct ocfs2_new_group_input)
 
 /*
  * Journal Flags (ocfs2_dinode.id1.journal1.i_flags)

fs/ocfs2/resize.c

@@ -356,7 +356,7 @@ int ocfs2_group_extend(struct inode * inode, int new_clusters)
 	}
 
 	mlog(0, "extend the last group at %llu, new clusters = %d\n",
-	     le64_to_cpu(group->bg_blkno), new_clusters);
+	     (unsigned long long)le64_to_cpu(group->bg_blkno), new_clusters);
 
 	handle = ocfs2_start_trans(osb, OCFS2_GROUP_EXTEND_CREDITS);
 	if (IS_ERR(handle)) {
@@ -396,3 +396,246 @@ out:
 	mlog_exit_void();
 	return ret;
 }
+
+static int ocfs2_check_new_group(struct inode *inode,
+				 struct ocfs2_dinode *di,
+				 struct ocfs2_new_group_input *input,
+				 struct buffer_head *group_bh)
+{
+	int ret;
+	struct ocfs2_group_desc *gd;
+	u16 cl_bpc = le16_to_cpu(di->id2.i_chain.cl_bpc);
+	unsigned int max_bits = le16_to_cpu(di->id2.i_chain.cl_cpg) *
+				le16_to_cpu(di->id2.i_chain.cl_bpc);
+
+
+	gd = (struct ocfs2_group_desc *)group_bh->b_data;
+
+	ret = -EIO;
+	if (!OCFS2_IS_VALID_GROUP_DESC(gd))
+		mlog(ML_ERROR, "Group descriptor # %llu isn't valid.\n",
+		     (unsigned long long)le64_to_cpu(gd->bg_blkno));
+	else if (di->i_blkno != gd->bg_parent_dinode)
+		mlog(ML_ERROR, "Group descriptor # %llu has bad parent "
+		     "pointer (%llu, expected %llu)\n",
+		     (unsigned long long)le64_to_cpu(gd->bg_blkno),
+		     (unsigned long long)le64_to_cpu(gd->bg_parent_dinode),
+		     (unsigned long long)le64_to_cpu(di->i_blkno));
+	else if (le16_to_cpu(gd->bg_bits) > max_bits)
+		mlog(ML_ERROR, "Group descriptor # %llu has bit count of %u\n",
+		     (unsigned long long)le64_to_cpu(gd->bg_blkno),
+		     le16_to_cpu(gd->bg_bits));
+	else if (le16_to_cpu(gd->bg_free_bits_count) > le16_to_cpu(gd->bg_bits))
+		mlog(ML_ERROR, "Group descriptor # %llu has bit count %u but "
+		     "claims that %u are free\n",
+		     (unsigned long long)le64_to_cpu(gd->bg_blkno),
+		     le16_to_cpu(gd->bg_bits),
+		     le16_to_cpu(gd->bg_free_bits_count));
+	else if (le16_to_cpu(gd->bg_bits) > (8 * le16_to_cpu(gd->bg_size)))
+		mlog(ML_ERROR, "Group descriptor # %llu has bit count %u but "
+		     "max bitmap bits of %u\n",
+		     (unsigned long long)le64_to_cpu(gd->bg_blkno),
+		     le16_to_cpu(gd->bg_bits),
+		     8 * le16_to_cpu(gd->bg_size));
+	else if (le16_to_cpu(gd->bg_chain) != input->chain)
+		mlog(ML_ERROR, "Group descriptor # %llu has bad chain %u "
+		     "while input has %u set.\n",
+		     (unsigned long long)le64_to_cpu(gd->bg_blkno),
+		     le16_to_cpu(gd->bg_chain), input->chain);
+	else if (le16_to_cpu(gd->bg_bits) != input->clusters * cl_bpc)
+		mlog(ML_ERROR, "Group descriptor # %llu has bit count %u but "
+		     "input has %u clusters set\n",
+		     (unsigned long long)le64_to_cpu(gd->bg_blkno),
+		     le16_to_cpu(gd->bg_bits), input->clusters);
+	else if (le16_to_cpu(gd->bg_free_bits_count) != input->frees * cl_bpc)
+		mlog(ML_ERROR, "Group descriptor # %llu has free bit count %u "
+		     "but it should have %u set\n",
+		     (unsigned long long)le64_to_cpu(gd->bg_blkno),
+		     le16_to_cpu(gd->bg_bits),
+		     input->frees * cl_bpc);
+	else
+		ret = 0;
+
+	return ret;
+}
+
+static int ocfs2_verify_group_and_input(struct inode *inode,
+					struct ocfs2_dinode *di,
+					struct ocfs2_new_group_input *input,
+					struct buffer_head *group_bh)
+{
+	u16 cl_count = le16_to_cpu(di->id2.i_chain.cl_count);
+	u16 cl_cpg = le16_to_cpu(di->id2.i_chain.cl_cpg);
+	u16 next_free = le16_to_cpu(di->id2.i_chain.cl_next_free_rec);
+	u32 cluster = ocfs2_blocks_to_clusters(inode->i_sb, input->group);
+	u32 total_clusters = le32_to_cpu(di->i_clusters);
+	int ret = -EINVAL;
+
+	if (cluster < total_clusters)
+		mlog(ML_ERROR, "add a group which is in the current volume.\n");
+	else if (input->chain >= cl_count)
+		mlog(ML_ERROR, "input chain exceeds the limit.\n");
+	else if (next_free != cl_count && next_free != input->chain)
+		mlog(ML_ERROR,
+		     "the add group should be in chain %u\n", next_free);
+	else if (total_clusters + input->clusters < total_clusters)
+		mlog(ML_ERROR, "add group's clusters overflow.\n");
+	else if (input->clusters > cl_cpg)
+		mlog(ML_ERROR, "the cluster exceeds the maximum of a group\n");
+	else if (input->frees > input->clusters)
+		mlog(ML_ERROR, "the free cluster exceeds the total clusters\n");
+	else if (total_clusters % cl_cpg != 0)
+		mlog(ML_ERROR,
+		     "the last group isn't full. Use group extend first.\n");
+	else if (input->group != ocfs2_which_cluster_group(inode, cluster))
+		mlog(ML_ERROR, "group blkno is invalid\n");
+	else if ((ret = ocfs2_check_new_group(inode, di, input, group_bh)))
+		mlog(ML_ERROR, "group descriptor check failed.\n");
+	else
+		ret = 0;
+
+	return ret;
+}
+
+/* Add a new group descriptor to global_bitmap. */
+int ocfs2_group_add(struct inode *inode, struct ocfs2_new_group_input *input)
+{
+	int ret;
+	handle_t *handle;
+	struct buffer_head *main_bm_bh = NULL;
+	struct inode *main_bm_inode = NULL;
+	struct ocfs2_dinode *fe = NULL;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	struct buffer_head *group_bh = NULL;
+	struct ocfs2_group_desc *group = NULL;
+	struct ocfs2_chain_list *cl;
+	struct ocfs2_chain_rec *cr;
+	u16 cl_bpc;
+
+	mlog_entry_void();
+
+	if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
+		return -EROFS;
+
+	main_bm_inode = ocfs2_get_system_file_inode(osb,
+						    GLOBAL_BITMAP_SYSTEM_INODE,
+						    OCFS2_INVALID_SLOT);
+	if (!main_bm_inode) {
+		ret = -EINVAL;
+		mlog_errno(ret);
+		goto out;
+	}
+
+	mutex_lock(&main_bm_inode->i_mutex);
+
+	ret = ocfs2_inode_lock(main_bm_inode, &main_bm_bh, 1);
+	if (ret < 0) {
+		mlog_errno(ret);
+		goto out_mutex;
+	}
+
+	fe = (struct ocfs2_dinode *)main_bm_bh->b_data;
+
+	if (le16_to_cpu(fe->id2.i_chain.cl_cpg) !=
+				 ocfs2_group_bitmap_size(osb->sb) * 8) {
+		mlog(ML_ERROR, "The disk is too old and small."
+		     " Force to do offline resize.");
+		ret = -EINVAL;
+		goto out_unlock;
+	}
+
+	ret = ocfs2_read_block(osb, input->group, &group_bh, 0, NULL);
+	if (ret < 0) {
+		mlog(ML_ERROR, "Can't read the group descriptor # %llu "
+		     "from the device.", (unsigned long long)input->group);
+		goto out_unlock;
+	}
+
+	ocfs2_set_new_buffer_uptodate(inode, group_bh);
+
+	ret = ocfs2_verify_group_and_input(main_bm_inode, fe, input, group_bh);
+	if (ret) {
+		mlog_errno(ret);
+		goto out_unlock;
+	}
+
+	mlog(0, "Add a new group  %llu in chain = %u, length = %u\n",
+	     (unsigned long long)input->group, input->chain, input->clusters);
+
+	handle = ocfs2_start_trans(osb, OCFS2_GROUP_ADD_CREDITS);
+	if (IS_ERR(handle)) {
+		mlog_errno(PTR_ERR(handle));
+		ret = -EINVAL;
+		goto out_unlock;
+	}
+
+	cl_bpc = le16_to_cpu(fe->id2.i_chain.cl_bpc);
+	cl = &fe->id2.i_chain;
+	cr = &cl->cl_recs[input->chain];
+
+	ret = ocfs2_journal_access(handle, main_bm_inode, group_bh,
+				   OCFS2_JOURNAL_ACCESS_WRITE);
+	if (ret < 0) {
+		mlog_errno(ret);
+		goto out_commit;
+	}
+
+	group = (struct ocfs2_group_desc *)group_bh->b_data;
+	group->bg_next_group = cr->c_blkno;
+
+	ret = ocfs2_journal_dirty(handle, group_bh);
+	if (ret < 0) {
+		mlog_errno(ret);
+		goto out_commit;
+	}
+
+	ret = ocfs2_journal_access(handle, main_bm_inode, main_bm_bh,
+				   OCFS2_JOURNAL_ACCESS_WRITE);
+	if (ret < 0) {
+		mlog_errno(ret);
+		goto out_commit;
+	}
+
+	if (input->chain == le16_to_cpu(cl->cl_next_free_rec)) {
+		le16_add_cpu(&cl->cl_next_free_rec, 1);
+		memset(cr, 0, sizeof(struct ocfs2_chain_rec));
+	}
+
+	cr->c_blkno = le64_to_cpu(input->group);
+	le32_add_cpu(&cr->c_total, input->clusters * cl_bpc);
+	le32_add_cpu(&cr->c_free, input->frees * cl_bpc);
+
+	le32_add_cpu(&fe->id1.bitmap1.i_total, input->clusters *cl_bpc);
+	le32_add_cpu(&fe->id1.bitmap1.i_used,
+		     (input->clusters - input->frees) * cl_bpc);
+	le32_add_cpu(&fe->i_clusters, input->clusters);
+
+	ocfs2_journal_dirty(handle, main_bm_bh);
+
+	spin_lock(&OCFS2_I(main_bm_inode)->ip_lock);
+	OCFS2_I(main_bm_inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
+	le64_add_cpu(&fe->i_size, input->clusters << osb->s_clustersize_bits);
+	spin_unlock(&OCFS2_I(main_bm_inode)->ip_lock);
+	i_size_write(main_bm_inode, le64_to_cpu(fe->i_size));
+
+	ocfs2_update_super_and_backups(main_bm_inode, input->clusters);
+
+out_commit:
+	ocfs2_commit_trans(osb, handle);
+out_unlock:
+	if (group_bh)
+		brelse(group_bh);
+
+	if (main_bm_bh)
+		brelse(main_bm_bh);
+
+	ocfs2_inode_unlock(main_bm_inode, 1);
+
+out_mutex:
+	mutex_unlock(&main_bm_inode->i_mutex);
+	iput(main_bm_inode);
+
+out:
+	mlog_exit_void();
+	return ret;
+}

fs/ocfs2/resize.h

@@ -27,5 +27,6 @@
 #define OCFS2_RESIZE_H
 
 int ocfs2_group_extend(struct inode * inode, int new_clusters);
+int ocfs2_group_add(struct inode *inode, struct ocfs2_new_group_input *input);
 
 #endif /* OCFS2_RESIZE_H */