Re: Patches for the patchqueue

[Mingming Cao <cmm@us.ibm.com>]

On Thu, 2008-06-05 at 15:25 +0530, Aneesh Kumar K.V wrote:
> f) clear the delay bit in ext4_da_get_block_write instead of
> __block_write_full_page
> so that we clear the delay bit for every successfull block allocation.
> We may fail
> while marking inode dirty in ext4_da_get_block_write after allocating
> block. So
> it is better to clear the delay bit in ext4_da_get_block_write rather
> than
> __block_write_full_page
> 

> Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
> ---

> @@ -1555,7 +1565,15 @@ static int ext4_da_get_block_write(struct inode
> *inode, sector_t iblock,
>                 bh_result->b_size = (ret << inode->i_blkbits);
> 
>                 /* release reserved-but-unused meta blocks */
> -               ext4_da_release_space(inode, ret, 0);
> +               if (buffer_delay(bh_result)) {
> +                       ext4_da_release_space(inode, ret, 0);
> +                       /*
> +                        * clear the delay bit now that we allocated
> +                        * blocks. If it is not a single block request
> +                        * we clear the delay bit in
> mpage_put_bnr_to_bhs
> +                        */
> +                       clear_buffer_delay(bh_result);
> +               }
> 
>                 /*
>                  * Update on-disk size along with block allocation

It seems with this fix, the buffer_delay bit is still cleared before the
ext4_mark_inode_dirty() could return error? Actually the already
allocated blocks are leaked if mark_inode-dirty() returns error, and we
cleared the buffer_delay for the buffer needs block.

Mingming