Re: raid5: handle_stripe_dirtying

[NeilBrown <neilb@suse.com>]

On Tue, 4 Aug 2015 18:32:37 +0000 Markus Stockhausen
<stockhausen@collogia.de> wrote:

> Hi Neil,
> 
> before sending a wrong patch. Could you help me to understand the reason 
> for an unconditional singular 

Sending a wrong patch is not such a bad thing - it would help me know
what you are thinking, and so reduce guess-work :-)

> 
> set_bit(STRIPE_HANDLE, &sh->state); 
> 
> in function handle_stripe_dirtying of raid5.c? It seems to be there since
> its introduction somewhere 8 years ago - patch "raid5: refactor handle_stripe5 
> and handle_stripe6 (v3)". 
> 
> If I understand the idea behind the flag right it is required to ensure 
> handling of the stripe in the next handle_stripe() run. That would only make
> sense if we set it unconditionally OR depending on some changes to the
> stripe. The above function does both. See a few lines below in the deep
> if-blocks.

So I'm guessing that you want to remove one of those?  Probably
justified.

That duplication goes back to
 Commit: 396a6123577d ("v2.4.5.4 -> v2.4.5.5")

I think your understanding of STRIPE_HANDLE is pretty spot-on.

NeilBrown

> 
> Thanks in advance.
> 
> Markus