This patch avoids doing an O(n**2)-complexity walk through every flex group.
Instead, it uses the already computed overhead information for the newly
allocated space, and simply adds it to the previously calculated
overhead stored in the superblock. This drastically reduces the time
taken to resize very large bigalloc filesystems (from 3+ hours for a
64TB fs down to milliseconds).
Signed-off-by: Oleg Kiselev <[email protected]>
---
v2:
- changed static function's name to be more descriptive
- removed useless cast
---
fs/ext4/resize.c | 23 +++++++++++++++++++++--
1 file changed, 21 insertions(+), 2 deletions(-)
diff --git a/fs/ext4/resize.c b/fs/ext4/resize.c
index 8b70a4701293..a69113b4ce4e 100644
--- a/fs/ext4/resize.c
+++ b/fs/ext4/resize.c
@@ -1380,6 +1380,17 @@ static int ext4_setup_new_descs(handle_t *handle, struct,
return err;
}
+static void ext4_add_overhead(struct super_block *sb,
+ const ext4_fsblk_t overhead)
+{
+ struct ext4_sb_info *sbi = EXT4_SB(sb);
+ struct ext4_super_block *es = sbi->s_es;
+
+ sbi->s_overhead += overhead;
+ es->s_overhead_clusters = cpu_to_le32(sbi->s_overhead);
+ smp_wmb();
+}
+
/*
* ext4_update_super() updates the super block so that the newly added
* groups can be seen by the filesystem.
@@ -1481,9 +1492,17 @@ static void ext4_update_super(struct super_block *sb,
}
/*
- * Update the fs overhead information
+ * Update the fs overhead information.
+ *
+ * For bigalloc, if the superblock already has a properly calculated
+ * overhead, update it with a value based on numbers already computed
+ * above for the newly allocated capacity.
*/
- ext4_calculate_overhead(sb);
+ if (ext4_has_feature_bigalloc(sb) && (sbi->s_overhead != 0))
+ ext4_add_overhead(sb,
+ EXT4_NUM_B2C(sbi, blocks_count - free_blocks));
+ else
+ ext4_calculate_overhead(sb);
if (test_opt(sb, DEBUG))
printk(KERN_DEBUG "EXT4-fs: added group %u:"
--
2.34.3
On Wed, 20 Jul 2022 04:26:22 +0000, Kiselev, Oleg wrote:
> This patch avoids doing an O(n**2)-complexity walk through every flex group.
> Instead, it uses the already computed overhead information for the newly
> allocated space, and simply adds it to the previously calculated
> overhead stored in the superblock. This drastically reduces the time
> taken to resize very large bigalloc filesystems (from 3+ hours for a
> 64TB fs down to milliseconds).
>
> [...]
Applied, thanks! (Some slight adjustments were needed to resolve a
merge conflict.)
[1/2] ext4: reduce computation of overhead during resize
commit: d985f3d81c0be1c673ac0462b26727b2a1aeb0d6
Best regards,
--
Theodore Ts'o <[email protected]>