On Mon, 02 Jul 2012 09:08:41 +0800 Shaohua Li <shli@kernel.org> wrote:

> Currently the sequential read detection is global wide. It's natural to make it
> per disk based, which can improve the detection for concurrent multiple
> sequential reads. And next patch will make SSD read balance not use distance
> based algorithm, where this change help detect truly sequential read for SSD.
> 
> Signed-off-by: Shaohua Li <shli@fusionio.com>
> ---
>  drivers/md/raid1.c |   29 ++++++-----------------------
>  drivers/md/raid1.h |   11 +++++------
>  2 files changed, 11 insertions(+), 29 deletions(-)
> 
> Index: linux/drivers/md/raid1.c
> ===================================================================
> --- linux.orig/drivers/md/raid1.c	2012-06-28 10:44:47.550666575 +0800
> +++ linux/drivers/md/raid1.c	2012-06-28 12:01:03.513137377 +0800
> @@ -483,7 +483,6 @@ static int read_balance(struct r1conf *c
>  	const sector_t this_sector = r1_bio->sector;
>  	int sectors;
>  	int best_good_sectors;
> -	int start_disk;
>  	int best_disk;
>  	int i;
>  	sector_t best_dist;
> @@ -503,20 +502,17 @@ static int read_balance(struct r1conf *c
>  	best_good_sectors = 0;
>  
>  	if (conf->mddev->recovery_cp < MaxSector &&
> -	    (this_sector + sectors >= conf->next_resync)) {
> +	    (this_sector + sectors >= conf->next_resync))
>  		choose_first = 1;
> -		start_disk = 0;
> -	} else {
> +	else
>  		choose_first = 0;
> -		start_disk = conf->last_used;
> -	}
>  
>  	for (i = 0 ; i < conf->raid_disks * 2 ; i++) {
>  		sector_t dist;
>  		sector_t first_bad;
>  		int bad_sectors;
>  
> -		int disk = start_disk + i;
> +		int disk = i;
>  		if (disk >= conf->raid_disks)
>  			disk -= conf->raid_disks;

One last little change please.  Could you make that loop:


     for (disk = 0; disk < conf->raid_disks * 2; disk++)

and remote the 'if' statement that wraps 'disk' back to zero
when it gets too big (which is buggy and has been fixed in my for-next).

With that change I'll take the patch.

Thanks,
NeilBrown