md: Export preferred I/O sizes and physical alignment

For each of the RAID levels set the preferred I/O hints according to
chunk size and stripe width.

Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

---

diff --git a/drivers/md/linear.c b/drivers/md/linear.c
--- a/drivers/md/linear.c
+++ b/drivers/md/linear.c
@@ -135,6 +135,17 @@ static linear_conf_t *linear_conf(mddev_
 
 		blk_queue_stack_limits(mddev->queue,
 				       rdev->bdev->bd_disk->queue);
+
+		if (rdev->raid_disk == 0) {
+			disk_topology_set(mddev->gendisk, 0, IOT_PHYS_OFFSET,
+					  bdev_phys_offset(rdev->bdev)
+					  + rdev->data_offset);
+			disk_topology_set(mddev->gendisk, 0, IOT_OPT_BLOCK_SIZE,
+					  bdev_optimal_io_block(rdev->bdev));
+			disk_topology_set(mddev->gendisk, 0, IOT_OPT_IO_LENGTH,
+					  bdev_optimal_io_length(rdev->bdev));
+		}
+
 		/* as we don't honour merge_bvec_fn, we must never risk
 		 * violating it, so limit ->max_sector to one PAGE, as
 		 * a one page request is never in violation.
diff --git a/drivers/md/md.c b/drivers/md/md.c
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -3325,6 +3325,7 @@ static struct kobject *md_probe(dev_t de
 	disk->private_data = mddev;
 	disk->queue = mddev->queue;
 	add_disk(disk);
+	disk_topology_register(disk);
 	mddev->gendisk = disk;
 	mutex_unlock(&disks_mutex);
 	error = kobject_init_and_add(&mddev->kobj, &md_ktype, &disk->dev.kobj,
@@ -6181,6 +6182,7 @@ static __exit void md_exit(void)
 		if (!disk)
 			continue;
 		export_array(mddev);
+		disk_topology_unregister(disk);
 		del_gendisk(disk);
 		put_disk(disk);
 		mddev->gendisk = NULL;
diff --git a/drivers/md/raid0.c b/drivers/md/raid0.c
--- a/drivers/md/raid0.c
+++ b/drivers/md/raid0.c
@@ -140,6 +140,17 @@ static int create_strip_zones (mddev_t *
 
 		blk_queue_stack_limits(mddev->queue,
 				       rdev1->bdev->bd_disk->queue);
+
+		if (rdev1->raid_disk == 0) {
+			disk_set_io_hints(mddev->gendisk,
+					  bdev_phys_offset(rdev1->bdev)
+					  + rdev1->data_offset,
+					  mddev->chunk_size,
+					  mddev->chunk_size * mddev->raid_disks);
+			printk(KERN_ERR "raid0: chunk_size=%u, disks=%u\n",
+			       mddev->chunk_size, mddev->raid_disks);
+		}
+
 		/* as we don't honour merge_bvec_fn, we must never risk
 		 * violating it, so limit ->max_sector to one PAGE, as
 		 * a one page request is never in violation.
diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c
--- a/drivers/md/raid1.c
+++ b/drivers/md/raid1.c
@@ -1970,6 +1970,14 @@ static int run(mddev_t *mddev)
 
 		blk_queue_stack_limits(mddev->queue,
 				       rdev->bdev->bd_disk->queue);
+
+		if (rdev->raid_disk == 0)
+			disk_set_io_hints(mddev->gendisk,
+					  bdev_phys_offset(rdev->bdev)
+					  + rdev->data_offset,
+					  bdev_optimal_io_block(rdev->bdev),
+					  bdev_optimal_io_length(rdev->bdev));
+
 		/* as we don't honour merge_bvec_fn, we must never risk
 		 * violating it, so limit ->max_sector to one PAGE, as
 		 * a one page request is never in violation.
diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
--- a/drivers/md/raid10.c
+++ b/drivers/md/raid10.c
@@ -2106,6 +2106,15 @@ static int run(mddev_t *mddev)
 
 		blk_queue_stack_limits(mddev->queue,
 				       rdev->bdev->bd_disk->queue);
+
+		if (rdev->raid_disk == 0)
+			disk_set_io_hints(mddev->gendisk,
+					  bdev_phys_offset(rdev->bdev)
+					  + rdev->data_offset,
+					  mddev->chunk_size,
+					  (mddev->chunk_size * mddev->raid_disks) 
+					  >> 1);
+
 		/* as we don't honour merge_bvec_fn, we must never risk
 		 * violating it, so limit ->max_sector to one PAGE, as
 		 * a one page request is never in violation.
diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -4368,6 +4368,19 @@ static int run(mddev_t *mddev)
 			       mdname(mddev));
 			goto abort;
 		}
+
+		if (rdev->raid_disk == 0) {
+			int disks = mddev->raid_disks - 1;
+
+			if (mddev->level == 6)
+				disks--;
+
+			disk_set_io_hints(mddev->gendisk,
+					  bdev_phys_offset(rdev->bdev)
+					  + rdev->data_offset,
+					  mddev->chunk_size,
+					  mddev->chunk_size * disks);
+		}
 	}
 
 	{