[Ocfs2-devel] [PATCH] ocfs2: fall back to buffer IO when append dio is disabled with file hole existing

[Changwei Ge]

On 2017/12/19 9:39, Joseph Qi wrote:
> 
> 
> On 17/12/19 09:27, Andrew Morton wrote:
>> On Tue, 19 Dec 2017 09:24:17 +0800 Joseph Qi <jiangqi903 at gmail.com> wrote:
>>
>>>> --- a/fs/ocfs2/aops.c~ocfs2-fall-back-to-buffer-io-when-append-dio-is-disabled-with-file-hole-existing-fix
>>>> +++ a/fs/ocfs2/aops.c
>>>> @@ -2469,10 +2469,9 @@ static int ocfs2_dio_end_io(struct kiocb
>>>>    * Will look for holes and unwritten extents in the range starting at
>>>>    * pos for count bytes (inclusive).
>>>>    */
>>>> -static int ocfs2_check_range_for_holes(struct inode *inode, loff_t pos,
>>>> -				       size_t count)
>>>> +static bool ocfs2_range_has_holes(struct inode *inode, loff_t pos, size_t count)
>>>>   {
>>>> -	int ret = 0;
>>>> +	bool ret = false;
>>>
>>> I have a different opinion here. If we change return value from int to
>>> bool, the error returned by ocfs2_get_clusters cannot be reflected. So
>>> I'd prefer the original version.
>>
>> But that error code is not used?
>>
> IMO, since ocfs2_get_clusters will get io involved, the caller shouldn't
> ignore its error.
> 
> Something like:
> ret = ocfs2_check_range_for_holes();
> if (ret < 0) {
> 	mlog_errno(ret);
> 	goto out;
> }
> 
> Then check if append dio feature has been enabled as well as write range
> and hole.

Accept this point.

> 
> Thanks,
> Joseph
>