md: Use new topology calls to indicate alignment and I/O sizes

Switch MD over to the new disk_stack_limits() function which checks for aligment and adjusts preferred I/O sizes when stacking. Also indicate preferred I/O sizes where applicable. Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com> Signed-off-by: NeilBrown <neilb@suse.de>

md: Use new topology calls to indicate alignment and I/O sizes
Switch MD over to the new disk_stack_limits() function which checks for aligment and adjusts preferred I/O sizes when stacking. Also indicate preferred I/O sizes where applicable. Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com> Signed-off-by: NeilBrown <neilb@suse.de>
Martin K. Petersen · NeilBrown
1 parent 5a4f13fad1
Showing 6 changed files with 39 additions and 19 deletions Side-by-side Diff
drivers/md/linear.c
drivers/md/multipath.c
drivers/md/raid0.c
drivers/md/raid1.c
drivers/md/raid10.c
drivers/md/raid5.c
@@ -166,8 +166,8 @@
 			rdev->sectors = sectors * mddev->chunk_sectors;
 		}
  
-		blk_queue_stack_limits(mddev->queue,
-				       rdev->bdev->bd_disk->queue);
+		disk_stack_limits(mddev->gendisk, rdev->bdev,
+				  rdev->data_offset << 9);
 		/* 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.
@@ -294,7 +294,8 @@
 	for (path = first; path <= last; path++)
 		if ((p=conf->multipaths+path)->rdev == NULL) {
 			q = rdev->bdev->bd_disk->queue;
-			blk_queue_stack_limits(mddev->queue, q);
+			disk_stack_limits(mddev->gendisk, rdev->bdev,
+					  rdev->data_offset << 9);
  
 		/* as we don't honour merge_bvec_fn, we must never risk
 		 * violating it, so limit ->max_sector to one PAGE, as
  
@@ -463,9 +464,9 @@
  
 		disk = conf->multipaths + disk_idx;
 		disk->rdev = rdev;
+		disk_stack_limits(mddev->gendisk, rdev->bdev,
+				  rdev->data_offset << 9);
  
-		blk_queue_stack_limits(mddev->queue,
-				       rdev->bdev->bd_disk->queue);
 		/* as we don't honour merge_bvec_fn, we must never risk
 		 * violating it, not that we ever expect a device with
 		 * a merge_bvec_fn to be involved in multipath */
@@ -170,8 +170,8 @@
 		}
 		dev[j] = rdev1;
  
-		blk_queue_stack_limits(mddev->queue,
-				       rdev1->bdev->bd_disk->queue);
+		disk_stack_limits(mddev->gendisk, rdev1->bdev,
+				  rdev1->data_offset << 9);
 		/* 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.
@@ -250,6 +250,11 @@
 		       mddev->chunk_sectors << 9);
 		goto abort;
 	}
+
+	blk_queue_io_min(mddev->queue, mddev->chunk_sectors << 9);
+	blk_queue_io_opt(mddev->queue,
+			 (mddev->chunk_sectors << 9) * mddev->raid_disks);
+
 	printk(KERN_INFO "raid0: done.\n");
 	mddev->private = conf;
 	return 0;
@@ -1123,8 +1123,8 @@
 	for (mirror = first; mirror <= last; mirror++)
 		if ( !(p=conf->mirrors+mirror)->rdev) {
  
-			blk_queue_stack_limits(mddev->queue,
-					       rdev->bdev->bd_disk->queue);
+			disk_stack_limits(mddev->gendisk, rdev->bdev,
+					  rdev->data_offset << 9);
 			/* 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.
@@ -1988,9 +1988,8 @@
 		disk = conf->mirrors + disk_idx;
  
 		disk->rdev = rdev;
-
-		blk_queue_stack_limits(mddev->queue,
-				       rdev->bdev->bd_disk->queue);
+		disk_stack_limits(mddev->gendisk, rdev->bdev,
+				  rdev->data_offset << 9);
 		/* 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.
@@ -1151,8 +1151,8 @@
 	for ( ; mirror <= last ; mirror++)
 		if ( !(p=conf->mirrors+mirror)->rdev) {
  
-			blk_queue_stack_limits(mddev->queue,
-					       rdev->bdev->bd_disk->queue);
+			disk_stack_limits(mddev->gendisk, rdev->bdev,
+					  rdev->data_offset << 9);
 			/* 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.
@@ -2044,7 +2044,7 @@
 static int run(mddev_t *mddev)
 {
 	conf_t *conf;
-	int i, disk_idx;
+	int i, disk_idx, chunk_size;
 	mirror_info_t *disk;
 	mdk_rdev_t *rdev;
 	int nc, fc, fo;
@@ -2130,6 +2130,14 @@
 	spin_lock_init(&conf->device_lock);
 	mddev->queue->queue_lock = &conf->device_lock;
  
+	chunk_size = mddev->chunk_sectors << 9;
+	blk_queue_io_min(mddev->queue, chunk_size);
+	if (conf->raid_disks % conf->near_copies)
+		blk_queue_io_opt(mddev->queue, chunk_size * conf->raid_disks);
+	else
+		blk_queue_io_opt(mddev->queue, chunk_size *
+				 (conf->raid_disks / conf->near_copies));
+
 	list_for_each_entry(rdev, &mddev->disks, same_set) {
 		disk_idx = rdev->raid_disk;
 		if (disk_idx >= mddev->raid_disks
@@ -2138,9 +2146,8 @@
 		disk = conf->mirrors + disk_idx;
  
 		disk->rdev = rdev;
-
-		blk_queue_stack_limits(mddev->queue,
-				       rdev->bdev->bd_disk->queue);
+		disk_stack_limits(mddev->gendisk, rdev->bdev,
+				  rdev->data_offset << 9);
 		/* 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.
@@ -4452,7 +4452,7 @@
 static int run(mddev_t *mddev)
 {
 	raid5_conf_t *conf;
-	int working_disks = 0;
+	int working_disks = 0, chunk_size;
 	mdk_rdev_t *rdev;
  
 	if (mddev->recovery_cp != MaxSector)
@@ -4607,6 +4607,14 @@
 	md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
  
 	blk_queue_merge_bvec(mddev->queue, raid5_mergeable_bvec);
+	chunk_size = mddev->chunk_sectors << 9;
+	blk_queue_io_min(mddev->queue, chunk_size);
+	blk_queue_io_opt(mddev->queue, chunk_size *
+			 (conf->raid_disks - conf->max_degraded));
+
+	list_for_each_entry(rdev, &mddev->disks, same_set)
+		disk_stack_limits(mddev->gendisk, rdev->bdev,
+				  rdev->data_offset << 9);
  
 	return 0;
 abort:
...	...	@@ -166,8 +166,8 @@
166	166	rdev->sectors = sectors * mddev->chunk_sectors;
167	167	}
168	168
169		- blk_queue_stack_limits(mddev->queue,
170		- rdev->bdev->bd_disk->queue);
	169	+ disk_stack_limits(mddev->gendisk, rdev->bdev,
	170	+ rdev->data_offset << 9);
171	171	/* as we don't honour merge_bvec_fn, we must never risk
172	172	* violating it, so limit ->max_sector to one PAGE, as
173	173	* a one page request is never in violation.
...	...	@@ -294,7 +294,8 @@
294	294	for (path = first; path <= last; path++)
295	295	if ((p=conf->multipaths+path)->rdev == NULL) {
296	296	q = rdev->bdev->bd_disk->queue;
297		- blk_queue_stack_limits(mddev->queue, q);
	297	+ disk_stack_limits(mddev->gendisk, rdev->bdev,
	298	+ rdev->data_offset << 9);
298	299
299	300	/* as we don't honour merge_bvec_fn, we must never risk
300	301	* violating it, so limit ->max_sector to one PAGE, as
301	302
...	...	@@ -463,9 +464,9 @@
463	464
464	465	disk = conf->multipaths + disk_idx;
465	466	disk->rdev = rdev;
	467	+ disk_stack_limits(mddev->gendisk, rdev->bdev,
	468	+ rdev->data_offset << 9);
466	469
467		- blk_queue_stack_limits(mddev->queue,
468		- rdev->bdev->bd_disk->queue);
469	470	/* as we don't honour merge_bvec_fn, we must never risk
470	471	* violating it, not that we ever expect a device with
471	472	* a merge_bvec_fn to be involved in multipath */
...	...	@@ -170,8 +170,8 @@
170	170	}
171	171	dev[j] = rdev1;
172	172
173		- blk_queue_stack_limits(mddev->queue,
174		- rdev1->bdev->bd_disk->queue);
	173	+ disk_stack_limits(mddev->gendisk, rdev1->bdev,
	174	+ rdev1->data_offset << 9);
175	175	/* as we don't honour merge_bvec_fn, we must never risk
176	176	* violating it, so limit ->max_sector to one PAGE, as
177	177	* a one page request is never in violation.
...	...	@@ -250,6 +250,11 @@
250	250	mddev->chunk_sectors << 9);
251	251	goto abort;
252	252	}
	253	+
	254	+ blk_queue_io_min(mddev->queue, mddev->chunk_sectors << 9);
	255	+ blk_queue_io_opt(mddev->queue,
	256	+ (mddev->chunk_sectors << 9) * mddev->raid_disks);
	257	+
253	258	printk(KERN_INFO "raid0: done.\n");
254	259	mddev->private = conf;
255	260	return 0;
...	...	@@ -1123,8 +1123,8 @@
1123	1123	for (mirror = first; mirror <= last; mirror++)
1124	1124	if ( !(p=conf->mirrors+mirror)->rdev) {
1125	1125
1126		- blk_queue_stack_limits(mddev->queue,
1127		- rdev->bdev->bd_disk->queue);
	1126	+ disk_stack_limits(mddev->gendisk, rdev->bdev,
	1127	+ rdev->data_offset << 9);
1128	1128	/* as we don't honour merge_bvec_fn, we must never risk
1129	1129	* violating it, so limit ->max_sector to one PAGE, as
1130	1130	* a one page request is never in violation.
...	...	@@ -1988,9 +1988,8 @@
1988	1988	disk = conf->mirrors + disk_idx;
1989	1989
1990	1990	disk->rdev = rdev;
1991		-
1992		- blk_queue_stack_limits(mddev->queue,
1993		- rdev->bdev->bd_disk->queue);
	1991	+ disk_stack_limits(mddev->gendisk, rdev->bdev,
	1992	+ rdev->data_offset << 9);
1994	1993	/* as we don't honour merge_bvec_fn, we must never risk
1995	1994	* violating it, so limit ->max_sector to one PAGE, as
1996	1995	* a one page request is never in violation.
...	...	@@ -1151,8 +1151,8 @@
1151	1151	for ( ; mirror <= last ; mirror++)
1152	1152	if ( !(p=conf->mirrors+mirror)->rdev) {
1153	1153
1154		- blk_queue_stack_limits(mddev->queue,
1155		- rdev->bdev->bd_disk->queue);
	1154	+ disk_stack_limits(mddev->gendisk, rdev->bdev,
	1155	+ rdev->data_offset << 9);
1156	1156	/* as we don't honour merge_bvec_fn, we must never risk
1157	1157	* violating it, so limit ->max_sector to one PAGE, as
1158	1158	* a one page request is never in violation.
...	...	@@ -2044,7 +2044,7 @@
2044	2044	static int run(mddev_t *mddev)
2045	2045	{
2046	2046	conf_t *conf;
2047		- int i, disk_idx;
	2047	+ int i, disk_idx, chunk_size;
2048	2048	mirror_info_t *disk;
2049	2049	mdk_rdev_t *rdev;
2050	2050	int nc, fc, fo;
...	...	@@ -2130,6 +2130,14 @@
2130	2130	spin_lock_init(&conf->device_lock);
2131	2131	mddev->queue->queue_lock = &conf->device_lock;
2132	2132
	2133	+ chunk_size = mddev->chunk_sectors << 9;
	2134	+ blk_queue_io_min(mddev->queue, chunk_size);
	2135	+ if (conf->raid_disks % conf->near_copies)
	2136	+ blk_queue_io_opt(mddev->queue, chunk_size * conf->raid_disks);
	2137	+ else
	2138	+ blk_queue_io_opt(mddev->queue, chunk_size *
	2139	+ (conf->raid_disks / conf->near_copies));
	2140	+
2133	2141	list_for_each_entry(rdev, &mddev->disks, same_set) {
2134	2142	disk_idx = rdev->raid_disk;
2135	2143	if (disk_idx >= mddev->raid_disks
...	...	@@ -2138,9 +2146,8 @@
2138	2146	disk = conf->mirrors + disk_idx;
2139	2147
2140	2148	disk->rdev = rdev;
2141		-
2142		- blk_queue_stack_limits(mddev->queue,
2143		- rdev->bdev->bd_disk->queue);
	2149	+ disk_stack_limits(mddev->gendisk, rdev->bdev,
	2150	+ rdev->data_offset << 9);
2144	2151	/* as we don't honour merge_bvec_fn, we must never risk
2145	2152	* violating it, so limit ->max_sector to one PAGE, as
2146	2153	* a one page request is never in violation.
...	...	@@ -4452,7 +4452,7 @@
4452	4452	static int run(mddev_t *mddev)
4453	4453	{
4454	4454	raid5_conf_t *conf;
4455		- int working_disks = 0;
	4455	+ int working_disks = 0, chunk_size;
4456	4456	mdk_rdev_t *rdev;
4457	4457
4458	4458	if (mddev->recovery_cp != MaxSector)
...	...	@@ -4607,6 +4607,14 @@
4607	4607	md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
4608	4608
4609	4609	blk_queue_merge_bvec(mddev->queue, raid5_mergeable_bvec);
	4610	+ chunk_size = mddev->chunk_sectors << 9;
	4611	+ blk_queue_io_min(mddev->queue, chunk_size);
	4612	+ blk_queue_io_opt(mddev->queue, chunk_size *
	4613	+ (conf->raid_disks - conf->max_degraded));
	4614	+
	4615	+ list_for_each_entry(rdev, &mddev->disks, same_set)
	4616	+ disk_stack_limits(mddev->gendisk, rdev->bdev,
	4617	+ rdev->data_offset << 9);
4610	4618
4611	4619	return 0;
4612	4620	abort: