Thanks Ted, I send a new fix. Please help review.
On 6/4/23 11:45, Theodore Ts'o wrote:
diff --git a/fs/ext4/block_validity.c b/fs/ext4/block_validity.c
index 5504f72bbbbe..5df357763975 100644
--- a/fs/ext4/block_validity.c
+++ b/fs/ext4/block_validity.c
@@ -224,11 +223,14 @@ int ext4_setup_system_zone(struct super_block *sb)
for (i=0; i < ngroups; i++) {
cond_resched();
- if (ext4_bg_has_super(sb, i) &&
- ((i < 5) || ((i % flex_size) == 0))) {
+ unsigned int sb_num = ext4_bg_has_super(sb, i);
+ unsigned long gdb_num = ext4_bg_num_gdb(sb, i);
+ unsigned int rsvd_gdt = le16_to_cpu(sbi->es->s_reserved_gdt_blocks);
+
+ if (sb_num != 0 || gdb_num != 0) {
ret = add_system_zone(system_blks,
ext4_group_first_block_no(sb, i),
- ext4_bg_num_gdb(sb, i) + 1, 0);
+ sb_num + gdb_num + rsvd_gdt, 0);
if (ret)
goto err;
}
How the reserved GDT blocks should be added to the system zone are not
handled correctly in this patch. It can't be unconditionally added to
all block groups.
See the logic in ext4_num_base_meta_clusters() in fs/ext4/balloc.c ---
without the EXT4_NUM_B2C() at the end of the function, since the
system zone tracking is done at the block level, not the cluster
level.
- Ted
