Re: [PATCH -next v5 05/32] ext4: set EXT4_MAP_NEW flag for delayed allocated blocks

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] The newly introduced EXT4_MAP_NEW flag is completely ignored by its only caller, rendering the patch ineffective (dead code) and failing to resolve the underlying bug.
--

--- Patch [5]: [PATCH -next v5 05/32] ext4: set EXT4_MAP_NEW flag for delayed allocated blocks ---
commit f71e7357d8a5fa9c83f4d4fc004cfa3640d60291
Author: Zhang Yi <[email protected]>

ext4: set EXT4_MAP_NEW flag for delayed allocated blocks

This commit sets the EXT4_MAP_NEW flag in ext4_da_map_blocks() to indicate
when a new delayed allocation block has been inserted. This aims to allow
callers to distinguish newly created delayed extents from existing ones.

Link: https://lore.kernel.org/linux-ext4/[email protected]/

> diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
> index 7601fe3618b1f..9dbece14ae569 100644
> --- a/fs/ext4/inode.c
> +++ b/fs/ext4/inode.c
> @@ -1990,7 +1990,7 @@ static int ext4_da_map_blocks(struct inode *inode, struct ext4_map_blocks *map)
>  		}
>  	}
>  
> -	map->m_flags |= EXT4_MAP_DELAYED;
> +	map->m_flags |= EXT4_MAP_DELAYED | EXT4_MAP_NEW;

[Severity: High]
Does this code actually allow callers to use the newly added flag?

Looking at ext4_da_get_block_prep(), which calls ext4_da_map_blocks(),
it unconditionally intercepts EXT4_MAP_DELAYED, sets BH_New, and returns
early without checking EXT4_MAP_NEW:

fs/ext4/inode.c:ext4_da_get_block_prep() {
    ...
	if (map.m_flags & EXT4_MAP_DELAYED) {
		map_bh(bh, inode->i_sb, invalid_block);
		set_buffer_new(bh);
		set_buffer_delay(bh);
		return 0;
	}
    ...
}

Because EXT4_MAP_DELAYED is always set alongside EXT4_MAP_NEW, this branch is
always taken, and set_buffer_new(bh) is unconditionally called. The code never
seems to evaluate map.m_flags & EXT4_MAP_NEW.

Is it possible this patch is incomplete and the underlying issue remains
unfixed because the new flag is ignored?

>  	retval = ext4_insert_delayed_blocks(inode, map->m_lblk, map->m_len);
>  	if (!retval)
>  		map->m_seq = READ_ONCE(EXT4_I(inode)->i_es_seq);

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=5