[PATCH] bcache/md: Use raid stripe size

[PATCH] bcache/md: Use raid stripe size

[Kent Overstreet]

Now that we've got code for raid5/6 stripe awareness, bcache just needs
to know about the stripes and when writing partial stripes is expensive
- we probably don't want to enable this optimization for raid1 or 10,
even though they have stripes. So add a flag to queue_limits.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
---
This is in my pile of bcache stuff queued up for 3.14 - if no one objects it'll
be in my bcache pull request:

 drivers/md/bcache/super.c | 6 ++++++
 drivers/md/raid5.c        | 1 +
 include/linux/blkdev.h    | 1 +
 3 files changed, 8 insertions(+)

diff --git a/drivers/md/bcache/super.c b/drivers/md/bcache/super.c
index 63ebef78df..e363efcf2b 100644
--- a/drivers/md/bcache/super.c
+++ b/drivers/md/bcache/super.c
@@ -1134,6 +1134,12 @@ static int cached_dev_init(struct cached_dev *dc, unsigned block_size)
 		hlist_add_head(&io->hash, dc->io_hash + RECENT_IO);
 	}
 
+	dc->disk.stripe_size = q->limits.io_opt >> 9;
+
+	if (dc->disk.stripe_size)
+		dc->partial_stripes_expensive =
+			q->limits.raid_partial_stripes_expensive;
+
 	ret = bcache_device_init(&dc->disk, block_size,
 			 dc->bdev->bd_part->nr_sects - dc->sb.data_offset);
 	if (ret)
diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index bef353c51c..4b5035f286 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -6104,6 +6104,7 @@ static int run(struct mddev *mddev)
 		blk_queue_io_min(mddev->queue, chunk_size);
 		blk_queue_io_opt(mddev->queue, chunk_size *
 				 (conf->raid_disks - conf->max_degraded));
+		mddev->queue->limits.raid_partial_stripes_expensive = 1;
 		/*
 		 * We can only discard a whole stripe. It doesn't make sense to
 		 * discard data disk but write parity disk
diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index 02cb6f0ea7..0375654adb 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -291,6 +291,7 @@ struct queue_limits {
 	unsigned char		discard_misaligned;
 	unsigned char		cluster;
 	unsigned char		discard_zeroes_data;
+	unsigned char		raid_partial_stripes_expensive;
 };
 
 struct request_queue {
-- 
1.8.5.1

Re: [PATCH] bcache/md: Use raid stripe size

[Mike Snitzer]

On Wed, Dec 18, 2013 at 8:12 PM, Kent Overstreet <kmo@daterainc.com> wrote:
> Now that we've got code for raid5/6 stripe awareness, bcache just needs
> to know about the stripes and when writing partial stripes is expensive
> - we probably don't want to enable this optimization for raid1 or 10,
> even though they have stripes. So add a flag to queue_limits.
>
> Signed-off-by: Kent Overstreet <kmo@daterainc.com>
> ---
> This is in my pile of bcache stuff queued up for 3.14 - if no one objects it'll
> be in my bcache pull request:
>
>  drivers/md/bcache/super.c | 6 ++++++
>  drivers/md/raid5.c        | 1 +
>  include/linux/blkdev.h    | 1 +
>  3 files changed, 8 insertions(+)

Hi Kent,

> diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
> index 02cb6f0ea7..0375654adb 100644
> --- a/include/linux/blkdev.h
> +++ b/include/linux/blkdev.h
> @@ -291,6 +291,7 @@ struct queue_limits {
>         unsigned char           discard_misaligned;
>         unsigned char           cluster;
>         unsigned char           discard_zeroes_data;
> +       unsigned char           raid_partial_stripes_expensive;
>  };
>
>  struct request_queue {

If you add a new member to queue_limits you need to update
blk_stack_limits() so that it propagates as you stack devices.

You probably also want to update blk_set_default_limits.

Mike

Re: [PATCH] bcache/md: Use raid stripe size

[Kent Overstreet]

On Mon, Jan 06, 2014 at 09:04:25PM -0500, Mike Snitzer wrote:
> On Wed, Dec 18, 2013 at 8:12 PM, Kent Overstreet <kmo@daterainc.com> wrote:
> > Now that we've got code for raid5/6 stripe awareness, bcache just needs
> > to know about the stripes and when writing partial stripes is expensive
> > - we probably don't want to enable this optimization for raid1 or 10,
> > even though they have stripes. So add a flag to queue_limits.
> >
> > Signed-off-by: Kent Overstreet <kmo@daterainc.com>
> > ---
> > This is in my pile of bcache stuff queued up for 3.14 - if no one objects it'll
> > be in my bcache pull request:
> >
> >  drivers/md/bcache/super.c | 6 ++++++
> >  drivers/md/raid5.c        | 1 +
> >  include/linux/blkdev.h    | 1 +
> >  3 files changed, 8 insertions(+)
> 
> Hi Kent,
> 
> > diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
> > index 02cb6f0ea7..0375654adb 100644
> > --- a/include/linux/blkdev.h
> > +++ b/include/linux/blkdev.h
> > @@ -291,6 +291,7 @@ struct queue_limits {
> >         unsigned char           discard_misaligned;
> >         unsigned char           cluster;
> >         unsigned char           discard_zeroes_data;
> > +       unsigned char           raid_partial_stripes_expensive;
> >  };
> >
> >  struct request_queue {
> 
> If you add a new member to queue_limits you need to update
> blk_stack_limits() so that it propagates as you stack devices.

Thanks, going to send Jens a new pull after he does some merging so I'll include
that in the updated version.