[Ocfs2-devel] [PATCH] ocfs2: don't evaluate buffer head to NULL managed by caller

[Joseph Qi]

On 18/3/30 10:17, Changwei Ge wrote:
>>>> Since we assume caller has to pass either all NULL or all non-NULL,
>>>> here we will only put bh internal allocated. Am I missing something?
>>> Thanks for your review.
>>> Yes, we will only put bh internally allocated.
>>> If bh is reserved in advance, we will not put it and re-assign it to NULL.
>>>
>> So this branch won't have risk, right?
> Sorry... I'm not sure if I understand you correctly.
> This branch will be walked through when previous part of bhs[] faces a read 
> failure in order to put bh allocated in ocfs2_read_blocks().
> And we assume all bh should be NULL or non-NULL, if new_bh is set, the back part 
> should also be put to release those buffer heads.
> 
> If I made a mistake or misunderstand you, please let me know.


I'm saying that sb_getblk() will only be called if bh hasn't been
allocated yet. That means if it fails, the bh to be put can be
guaranteed internal allocated.
Also I don't think the WARN check is necessary as this is common path
and will bring additional cpu consumption. We can make it clear at
comments of ocfs2_read_blocks() that either all NULL or non-NULL bhs
is prerequisite for the caller. And then we make sure we won't put bh
that is allocated outside.

Thanks,
Joseph