Re: [PATCH] FAT: optimize FSINFO writeback

[OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>]

steve@digidescorp.com writes:

> Only write the FSINFO block back to disk when its contents change.
> This optimization can be important when the underlying physical media
> can wear out, i.e. Flash.

I have no objection to this though.  Was this tested on recent version?
Well, now, we are using sb->s_dirty for fsinfo, so I'm wondering why
this happen frequently.

Thanks.

> Signed-off-by: Steven J. Magnani <steve@digidescorp.com>
> ---
> diff -uprN a/fs/fat/misc.c b/fs/fat/misc.c
> --- a/fs/fat/misc.c	2009-06-29 11:12:40.000000000 -0500
> +++ b/fs/fat/misc.c	2009-06-29 11:46:45.000000000 -0500
> @@ -61,11 +61,25 @@ void fat_clusters_flush(struct super_blo
>  		       le32_to_cpu(fsinfo->signature2),
>  		       sbi->fsinfo_sector);
>  	} else {
> -		if (sbi->free_clusters != -1)
> -			fsinfo->free_clusters = cpu_to_le32(sbi->free_clusters);
> -		if (sbi->prev_free != -1)
> -			fsinfo->next_cluster = cpu_to_le32(sbi->prev_free);
> -		mark_buffer_dirty(bh);
> +		char write_needed = 0;
> +		__le32 le_value;
> +
> +		if (sbi->free_clusters != -1) {
> +			le_value = cpu_to_le32(sbi->free_clusters);
> +			if (fsinfo->free_clusters != le_value) {
> +				fsinfo->free_clusters = le_value;
> +				write_needed = 1;
> +			}
> +		}
> +		if (sbi->prev_free != -1) {
> +			le_value = cpu_to_le32(sbi->prev_free);
> +			if (fsinfo->next_cluster != le_value) {
> +				fsinfo->next_cluster = le_value;
> +				write_needed = 1;
> +			}
> +		}
> +		if (write_needed)
> +			mark_buffer_dirty(bh);
>  	}
>  	brelse(bh);
>  }
>

-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>