Re: [PATCH v3 03/25] dcache: Enable lockless update of d_count in dentry structure

[Waiman Long <waiman.long@hp.com>]

On 07/03/2013 04:37 PM, Linus Torvalds wrote:
> This patch grew a lot, and that seems to be mainly because of bad reasons.

That is the main reason why I choose to implement it the way it was in 
my previous version. As I add one more level to access d_lock and 
d_count, I need to change a lot more files.

> I'd suggest dropping the whole
> "lockref_ret_count()"/"lockref_ret_lock()" helpers, which cause all
> the annoyance, and just make people use the members directly.

Yes, I can do that. They are used in not that many places.

> Then, just do
>
>     #define d_lock d_lockref.lockref_lock
>
> or similar, so that all the existing code just continues to work,
> without the need for the syntactic changes:
>
> -       spin_lock(&dentry->d_lock);
> +       d_lock(dentry);

I had been thinking about that. The use of d_lock should be pretty safe 
as I didn't see that variable name used in other places. I didn't do it 
because I am afraid that people may say that using macro mapping like 
this is not a good idea. By doing that, the patch should shrink 
considerably.

> For d_count, we probably do need to have the wrapper macro:
>
>    #define dentry_count(dentry)  ((dentry)->d_lockref.lockref_count)
>
> and change the existing users of "dentry->d_count" to use that, but
> there are fewer of those than there are of people taking the dentry
> lock. And most of them are in fs/dcache.c and would be affected by
> this set of patches anyway.

The d_count name is not unique to the dentry structure. So files that 
access d_count have to be modified explicitly.

I will see if there are more feedback and send an updated patchset by 
the end of this week or early next week.

Regards,
Longman