Re: Is this a bug?

[Amir Goldstein <amir73il@gmail.com>]

On Sun, Apr 3, 2011 at 7:51 AM, Yongqiang Yang <xiaoqiangnk@gmail.com> wrote:
> On Sun, Apr 3, 2011 at 5:24 PM, Ding Dinghua <dingdinghua85@gmail.com> wrote:
>> 2011/4/3 Amir Goldstein <amir73il@gmail.com>:
>>> On Sat, Apr 2, 2011 at 1:05 AM, Ding Dinghua <dingdinghua85@gmail.com> wrote:
>>>> When truncate files and free blocks, the following codes make me puzzled:
>>>>
>>>> void ext4_free_blocks(handle_t *handle, struct inode *inode,
>>>>                      struct buffer_head *bh, ext4_fsblk_t block,
>>>>                      unsigned long count, int flags)
>>>> {
>>>>        if (flags & EXT4_FREE_BLOCKS_FORGET) {
>>>>                struct buffer_head *tbh = bh;
>>>>                int i;
>>>>
>>>>                BUG_ON(bh && (count > 1));
>>>>
>>>>                for (i = 0; i < count; i++) {
>>>>                        if (!bh)
>>>>                                tbh = sb_find_get_block(inode->i_sb,
>>>>                                                        block + i);
>>>>                        if (unlikely(!tbh))
>>>>                                continue;
>>>>                        ext4_forget(handle, flags & EXT4_FREE_BLOCKS_METADATA,
>>>>                                    inode, tbh, block + i);
>>>>                }
>>>>        }
>>>> }
>>>>
>>>> I notice that ext4_forget mainly do two things:
>>>>    a)  call jbd2_journa_forget to forget the modification of some buffer head
>>>>    b)  or deal with the revoke issue
>>>> however, if we are freeing data blocks && ext4_forget get into branch a),
>>>
>>> Simple. we don't pass the FORGET flag when freeing data blocks,
>>> only when freeing metadata blocks, which may have been journalled
>>> already.
>>> I am not sure about the journal=data case through.
>> Thanks for reply. The reason for me to dip into this issue is journal=data mode.
> I think data processing in journal=data mode is similar to metadata
> processing in joutnal=ordered mode.

not exactly. Ding is right to be puzzled, because journal=data
journals inode page cache buffers and not block device buffers, like
metadata.
However, I think what happens is that prior to ext4_clear_blocks, which
invokes ext4_free_blocks with the FORGET flag, ext4_invalidatepage
will have been called already and forget about data buffers which were modified
in the current running transaction.
ext4_free_blocks will then handle revoke of data blocks which were modified in
previous transaction. for example, in orphan cleanup, ext4_free_blocks will be
will be called from ext4_truncate, but there will be no pages to invalidate.

this is my guess. I haven't studied this case thoroughly.

>>>
>>>> tbh is not the buffer_head which journal took care of in ext4_write_begin,
>>>> so i'm puzzled with this.
>>>>
>>>> Could anyone explain it to me? Thanks.
>>>>
>>>> --
>>>> Ding Dinghua
>>>> --
>>>> To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
>>>> the body of a message to majordomo@vger.kernel.org
>>>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>>>>
>>>
>>
>>
>>
>> --
>> Ding Dinghua
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>>
>
>
>
> --
> Best Wishes
> Yongqiang Yang
>
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html