Re: [PATCH][RFC] RAID5/DMA/memcpy: zero copy the bio page when possible

[NeilBrown <neilb@suse.de>]

[-- Attachment #1: Type: text/plain, Size: 1349 bytes --]

On Wed, 28 Dec 2011 13:20:08 +0800 <b29237@freescale.com> wrote:

> From: Forrest shi <b29237@freescale.com>
> 
> 	use bio page directly instead of copying the bio page to stripe
> 	header page when possible.
> 
> 	Signed-off-by: Forrest Shi<b29237@freescale.com>

I'm sorry but if you want anyone to review this and comment on it you need to
give us a bit more help - like telling us what the purpose of the patch is.

There are a couple of hints but what is needed is a 'big picture'
description.  What is the problem you are fixing?  How are you fixing it?
What are the costs/benefits?

> @@ -754,10 +764,16 @@ ops_run_compute5(struct stripe_head *sh, struct raid5_percpu *percpu)
>  		__func__, (unsigned long long)sh->sector, target);
>  	BUG_ON(!test_bit(R5_Wantcompute, &tgt->flags));
>  
> -	for (i = disks; i--; )
> +	for (i = disks; i--; ) {
> +		struct r5dev *dev = &sh->dev[i];
> +		struct page *pg = dev->page;
>  		if (i != target)
> -			xor_srcs[count++] = sh->dev[i].page;
> -
> +#ifdef CONFIG_OPTIMIZE_FSL_DMA_MEMCPY
> +			if (test_bit(R5_DirectAccess, &dev->flags))
> +				pg = dev->req.bi_io_vec[0].bv_page;
> +#endif
> +			xor_srcs[count++] = pg;
> +	}
>  	atomic_inc(&sh->count);

And something is very wrong here.  I think you need { } around the body of
the 'if' statement.

NeilBrown


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 828 bytes --]