[dm-devel] Re: [2.6.22 PATCH 01/26] dm: merge max_hw_sector

[Neil Brown]

On Tuesday May 8, agk at redhat.com wrote:
> From: Neil Brown <neilb at suse.de>
> 
> Make sure dm honours max_hw_sectors of underlying devices

I think I want to un-ack this patch (if that makes sense).

I don't think max_hw_sectors is at all relevant to dm or md.

I looked back at the mail thread that sparked this.  Someone reported
a problem.  I suggested this patch based on differences between
combine_restrictions_low and blk_queue_stack_limits, and we never
heard anything more.

As far as I can tell, max_hw_sectors is only used by bio_add_pc_page
which is only used for the scsi SG_IO ioctl.  i.e. it is only needed
when bypassing the block layer and talking very directly to a low
level device.
(Did I get that right Jens?)

So I'm sure this patch doesn't fix the original problem, and I don't
think we really want this patch (it won't hurt, except that it might
confuse readers).  We should probably remove
	t->max_hw_sectors = min_not_zero(t->max_hw_sectors,b->max_hw_sectors);
from blk_queue_stack_limits.

NeilBrown



> 
> Signed-off-by: Neil Brown <neilb at suse.de>
> Signed-off-by: Alasdair G Kergon <agk at redhat.com>
> ---
> 
> ### Diffstat output
>  drivers/md/dm-table.c         |    9 +++++++++
>  include/linux/device-mapper.h |    1 +
>  2 files changed, 10 insertions(+)
> 
> Index: linux-2.6.21/drivers/md/dm-table.c
> ===================================================================
> --- linux-2.6.21.orig/drivers/md/dm-table.c	2007-05-01 17:38:32.000000000 +0100
> +++ linux-2.6.21/drivers/md/dm-table.c	2007-05-01 17:40:45.000000000 +0100
> @@ -99,6 +99,9 @@ static void combine_restrictions_low(str
>  	lhs->max_segment_size =
>  		min_not_zero(lhs->max_segment_size, rhs->max_segment_size);
>  
> +	lhs->max_hw_sectors =
> +		min_not_zero(lhs->max_hw_sectors, rhs->max_hw_sectors);
> +
>  	lhs->seg_boundary_mask =
>  		min_not_zero(lhs->seg_boundary_mask, rhs->seg_boundary_mask);
>  
> @@ -561,6 +564,9 @@ void dm_set_device_limits(struct dm_targ
>  	rs->max_segment_size =
>  		min_not_zero(rs->max_segment_size, q->max_segment_size);
>  
> +	rs->max_hw_sectors =
> +		min_not_zero(rs->max_hw_sectors, q->max_hw_sectors);
> +
>  	rs->seg_boundary_mask =
>  		min_not_zero(rs->seg_boundary_mask,
>  			     q->seg_boundary_mask);
> @@ -696,6 +702,8 @@ static void check_for_valid_limits(struc
>  {
>  	if (!rs->max_sectors)
>  		rs->max_sectors = SAFE_MAX_SECTORS;
> +	if (!rs->max_hw_sectors)
> +		rs->max_hw_sectors = SAFE_MAX_SECTORS;
>  	if (!rs->max_phys_segments)
>  		rs->max_phys_segments = MAX_PHYS_SEGMENTS;
>  	if (!rs->max_hw_segments)
> @@ -889,6 +897,7 @@ void dm_table_set_restrictions(struct dm
>  	q->max_hw_segments = t->limits.max_hw_segments;
>  	q->hardsect_size = t->limits.hardsect_size;
>  	q->max_segment_size = t->limits.max_segment_size;
> +	q->max_hw_sectors = t->limits.max_hw_sectors;
>  	q->seg_boundary_mask = t->limits.seg_boundary_mask;
>  	if (t->limits.no_cluster)
>  		q->queue_flags &= ~(1 << QUEUE_FLAG_CLUSTER);
> Index: linux-2.6.21/include/linux/device-mapper.h
> ===================================================================
> --- linux-2.6.21.orig/include/linux/device-mapper.h	2007-05-01 17:38:32.000000000 +0100
> +++ linux-2.6.21/include/linux/device-mapper.h	2007-05-01 17:40:45.000000000 +0100
> @@ -115,6 +115,7 @@ struct io_restrictions {
>  	unsigned short		max_hw_segments;
>  	unsigned short		hardsect_size;
>  	unsigned int		max_segment_size;
> +	unsigned int		max_hw_sectors;
>  	unsigned long		seg_boundary_mask;
>  	unsigned char		no_cluster; /* inverted so that 0 is default */
>  };