[PATCH U-BOOT 2/3] btrfs: btrfs_file_read: allow opportunistic read until the end

[Qu Wenruo]

On 2023/4/18 09:17, Dominique Martinet wrote:
> From: Dominique Martinet <dominique.martinet at atmark-techno.com>
> 
> btrfs_file_read main 'aligned read' loop would limit the last read to
> the aligned end even if the data is present in the extent: there is no
> reason not to read the data that we know will be presented correctly.
> 
> If that somehow fails cur only advances up to the aligned_end anyway and
> the file tail is read through the old code unchanged; this could be
> required if e.g. the end data is inlined.
> 
> Signed-off-by: Dominique Martinet <dominique.martinet at atmark-techno.com>
> ---
>   fs/btrfs/inode.c | 17 +++++++++--------
>   1 file changed, 9 insertions(+), 8 deletions(-)
> 
> diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
> index 3d6e39e6544d..efffec0f2e68 100644
> --- a/fs/btrfs/inode.c
> +++ b/fs/btrfs/inode.c
> @@ -663,7 +663,8 @@ int btrfs_file_read(struct btrfs_root *root, u64 ino, u64 file_offset, u64 len,
>   	struct btrfs_path path;
>   	struct btrfs_key key;
>   	u64 aligned_start = round_down(file_offset, fs_info->sectorsize);
> -	u64 aligned_end = round_down(file_offset + len, fs_info->sectorsize);
> +	u64 end = file_offset + len;
> +	u64 aligned_end = round_down(end, fs_info->sectorsize);
>   	u64 next_offset;
>   	u64 cur = aligned_start;
>   	int ret = 0;
> @@ -743,26 +744,26 @@ int btrfs_file_read(struct btrfs_root *root, u64 ino, u64 file_offset, u64 len,
>   		extent_num_bytes = btrfs_file_extent_num_bytes(path.nodes[0],
>   							       fi);
>   		ret = btrfs_read_extent_reg(&path, fi, cur,
> -				min(extent_num_bytes, aligned_end - cur),
> +				min(extent_num_bytes, end - cur),
>   				dest + cur - file_offset);
>   		if (ret < 0)
>   			goto out;
> -		cur += min(extent_num_bytes, aligned_end - cur);
> +		cur += min(extent_num_bytes, end - cur);
>   	}
>   
>   	/* Read the tailing unaligned part*/

Can we remove this part completely?

IIRC if we read until the target end, the unaligned end part can be 
completely removed then.

Thanks,
Qu
> -	if (file_offset + len != aligned_end) {
> +	if (file_offset + len != cur) {
>   		btrfs_release_path(&path);
> -		ret = lookup_data_extent(root, &path, ino, aligned_end,
> +		ret = lookup_data_extent(root, &path, ino, cur,
>   					 &next_offset);
>   		/* <0 is error, >0 means no extent */
>   		if (ret)
>   			goto out;
>   		fi = btrfs_item_ptr(path.nodes[0], path.slots[0],
>   				    struct btrfs_file_extent_item);
> -		ret = read_and_truncate_page(&path, fi, aligned_end,
> -				file_offset + len - aligned_end,
> -				dest + aligned_end - file_offset);
> +		ret = read_and_truncate_page(&path, fi, cur,
> +				file_offset + len - cur,
> +				dest + cur - file_offset);
>   	}
>   out:
>   	btrfs_release_path(&path);
>