block: Require subsystems to explicitly allocate bio_set integrity mempool

MD and DM create a new bio_set for every metadevice. Each bio_set has an integrity mempool attached regardless of whether the metadevice is capable of passing integrity metadata. This is a waste of memory. Instead we defer the allocation decision to MD and DM since we know at metadevice creation time whether integrity passthrough is needed or not. Automatic integrity mempool allocation can then be removed from bioset_create() and we make an explicit integrity allocation for the fs_bio_set. Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Reported-by: Zdenek Kabelac <zkabelac@redhat.com> Acked-by: Mike Snitzer <snizer@redhat.com> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>

block: Require subsystems to explicitly allocate bio_set integrity mempool
MD and DM create a new bio_set for every metadevice. Each bio_set has an integrity mempool attached regardless of whether the metadevice is capable of passing integrity metadata. This is a waste of memory. Instead we defer the allocation decision to MD and DM since we know at metadevice creation time whether integrity passthrough is needed or not. Automatic integrity mempool allocation can then be removed from bioset_create() and we make an explicit integrity allocation for the fs_bio_set. Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Reported-by: Zdenek Kabelac <zkabelac@redhat.com> Acked-by: Mike Snitzer <snizer@redhat.com> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
Martin K. Petersen · Jens Axboe
1 parent 82f04ab47e
Showing 11 changed files with 41 additions and 22 deletions Side-by-side Diff
drivers/md/dm-table.c
drivers/md/dm.c
drivers/md/dm.h
drivers/md/linear.c
drivers/md/md.c
drivers/md/multipath.c
drivers/md/raid0.c
drivers/md/raid1.c
drivers/md/raid10.c
fs/bio-integrity.c
fs/bio.c
@@ -55,6 +55,7 @@
 	struct dm_target *targets;
  
 	unsigned discards_supported:1;
+	unsigned integrity_supported:1;
  
 	/*
 	 * Indicates the rw permissions for the new logical
@@ -859,7 +860,7 @@
 		return -EINVAL;
 	}
  
-	t->mempools = dm_alloc_md_mempools(type);
+	t->mempools = dm_alloc_md_mempools(type, t->integrity_supported);
 	if (!t->mempools)
 		return -ENOMEM;
  
  
@@ -935,8 +936,10 @@
 	struct dm_dev_internal *dd;
  
 	list_for_each_entry(dd, devices, list)
-		if (bdev_get_integrity(dd->dm_dev.bdev))
+		if (bdev_get_integrity(dd->dm_dev.bdev)) {
+			t->integrity_supported = 1;
 			return blk_integrity_register(dm_disk(md), NULL);
+		}
  
 	return 0;
 }
@@ -2620,9 +2620,10 @@
 }
 EXPORT_SYMBOL_GPL(dm_noflush_suspending);
  
-struct dm_md_mempools *dm_alloc_md_mempools(unsigned type)
+struct dm_md_mempools *dm_alloc_md_mempools(unsigned type, unsigned integrity)
 {
 	struct dm_md_mempools *pools = kmalloc(sizeof(*pools), GFP_KERNEL);
+	unsigned int pool_size = (type == DM_TYPE_BIO_BASED) ? 16 : MIN_IOS;
  
 	if (!pools)
 		return NULL;
  
  
@@ -2639,12 +2640,17 @@
 	if (!pools->tio_pool)
 		goto free_io_pool_and_out;
  
-	pools->bs = (type == DM_TYPE_BIO_BASED) ?
-		    bioset_create(16, 0) : bioset_create(MIN_IOS, 0);
+	pools->bs = bioset_create(pool_size, 0);
 	if (!pools->bs)
 		goto free_tio_pool_and_out;
  
+	if (integrity && bioset_integrity_create(pools->bs, pool_size))
+		goto free_bioset_and_out;
+
 	return pools;
+
+free_bioset_and_out:
+	bioset_free(pools->bs);
  
 free_tio_pool_and_out:
 	mempool_destroy(pools->tio_pool);
@@ -149,7 +149,7 @@
 /*
  * Mempool operations
  */
-struct dm_md_mempools *dm_alloc_md_mempools(unsigned type);
+struct dm_md_mempools *dm_alloc_md_mempools(unsigned type, unsigned integrity);
 void dm_free_md_mempools(struct dm_md_mempools *pools);
  
 #endif
@@ -210,8 +210,7 @@
 	blk_queue_merge_bvec(mddev->queue, linear_mergeable_bvec);
 	mddev->queue->backing_dev_info.congested_fn = linear_congested;
 	mddev->queue->backing_dev_info.congested_data = mddev;
-	md_integrity_register(mddev);
-	return 0;
+	return md_integrity_register(mddev);
 }
  
 static void free_conf(struct rcu_head *head)
@@ -1803,8 +1803,12 @@
 			mdname(mddev));
 		return -EINVAL;
 	}
-	printk(KERN_NOTICE "md: data integrity on %s enabled\n",
-		mdname(mddev));
+	printk(KERN_NOTICE "md: data integrity enabled on %s\n", mdname(mddev));
+	if (bioset_integrity_create(mddev->bio_set, BIO_POOL_SIZE)) {
+		printk(KERN_ERR "md: failed to create integrity pool for %s\n",
+		       mdname(mddev));
+		return -EINVAL;
+	}
 	return 0;
 }
 EXPORT_SYMBOL(md_integrity_register);
@@ -315,7 +315,7 @@
 			p->rdev = rdev;
 			goto abort;
 		}
-		md_integrity_register(mddev);
+		err = md_integrity_register(mddev);
 	}
 abort:
  
@@ -489,7 +489,10 @@
  
 	mddev->queue->backing_dev_info.congested_fn = multipath_congested;
 	mddev->queue->backing_dev_info.congested_data = mddev;
-	md_integrity_register(mddev);
+
+	if (md_integrity_register(mddev))
+		goto out_free_conf;
+
 	return 0;
  
 out_free_conf:
@@ -379,8 +379,7 @@
  
 	blk_queue_merge_bvec(mddev->queue, raid0_mergeable_bvec);
 	dump_zones(mddev);
-	md_integrity_register(mddev);
-	return 0;
+	return md_integrity_register(mddev);
 }
  
 static int raid0_stop(mddev_t *mddev)
@@ -1132,7 +1132,7 @@
 			p->rdev = rdev;
 			goto abort;
 		}
-		md_integrity_register(mddev);
+		err = md_integrity_register(mddev);
 	}
 abort:
  
@@ -2017,8 +2017,7 @@
  
 	mddev->queue->backing_dev_info.congested_fn = raid1_congested;
 	mddev->queue->backing_dev_info.congested_data = mddev;
-	md_integrity_register(mddev);
-	return 0;
+	return md_integrity_register(mddev);
 }
  
 static int stop(mddev_t *mddev)
@@ -1188,7 +1188,7 @@
 			p->rdev = rdev;
 			goto abort;
 		}
-		md_integrity_register(mddev);
+		err = md_integrity_register(mddev);
 	}
 abort:
  
@@ -2343,7 +2343,10 @@
  
 	if (conf->near_copies < conf->raid_disks)
 		blk_queue_merge_bvec(mddev->queue, raid10_mergeable_bvec);
-	md_integrity_register(mddev);
+
+	if (md_integrity_register(mddev))
+		goto out_free_conf;
+
 	return 0;
  
 out_free_conf:
@@ -761,6 +761,9 @@
 {
 	unsigned int max_slab = vecs_to_idx(BIO_MAX_PAGES);
  
+	if (bs->bio_integrity_pool)
+		return 0;
+
 	bs->bio_integrity_pool =
 		mempool_create_slab_pool(pool_size, bip_slab[max_slab].slab);
  
@@ -1636,9 +1636,6 @@
 	if (!bs->bio_pool)
 		goto bad;
  
-	if (bioset_integrity_create(bs, pool_size))
-		goto bad;
-
 	if (!biovec_create_pools(bs, pool_size))
 		return bs;
  
@@ -1681,6 +1678,9 @@
 	fs_bio_set = bioset_create(BIO_POOL_SIZE, 0);
 	if (!fs_bio_set)
 		panic("bio: can't allocate bios\n");
+
+	if (bioset_integrity_create(fs_bio_set, BIO_POOL_SIZE))
+		panic("bio: can't create integrity pool\n");
  
 	bio_split_pool = mempool_create_kmalloc_pool(BIO_SPLIT_ENTRIES,
 						     sizeof(struct bio_pair));
...	...	@@ -55,6 +55,7 @@
55	55	struct dm_target *targets;
56	56
57	57	unsigned discards_supported:1;
	58	+ unsigned integrity_supported:1;
58	59
59	60	/*
60	61	* Indicates the rw permissions for the new logical
...	...	@@ -859,7 +860,7 @@
859	860	return -EINVAL;
860	861	}
861	862
862		- t->mempools = dm_alloc_md_mempools(type);
	863	+ t->mempools = dm_alloc_md_mempools(type, t->integrity_supported);
863	864	if (!t->mempools)
864	865	return -ENOMEM;
865	866
866	867
...	...	@@ -935,8 +936,10 @@
935	936	struct dm_dev_internal *dd;
936	937
937	938	list_for_each_entry(dd, devices, list)
938		- if (bdev_get_integrity(dd->dm_dev.bdev))
	939	+ if (bdev_get_integrity(dd->dm_dev.bdev)) {
	940	+ t->integrity_supported = 1;
939	941	return blk_integrity_register(dm_disk(md), NULL);
	942	+ }
940	943
941	944	return 0;
942	945	}
...	...	@@ -2620,9 +2620,10 @@
2620	2620	}
2621	2621	EXPORT_SYMBOL_GPL(dm_noflush_suspending);
2622	2622
2623		-struct dm_md_mempools *dm_alloc_md_mempools(unsigned type)
	2623	+struct dm_md_mempools *dm_alloc_md_mempools(unsigned type, unsigned integrity)
2624	2624	{
2625	2625	struct dm_md_mempools pools = kmalloc(sizeof(pools), GFP_KERNEL);
	2626	+ unsigned int pool_size = (type == DM_TYPE_BIO_BASED) ? 16 : MIN_IOS;
2626	2627
2627	2628	if (!pools)
2628	2629	return NULL;
2629	2630
2630	2631
...	...	@@ -2639,12 +2640,17 @@
2639	2640	if (!pools->tio_pool)
2640	2641	goto free_io_pool_and_out;
2641	2642
2642		- pools->bs = (type == DM_TYPE_BIO_BASED) ?
2643		- bioset_create(16, 0) : bioset_create(MIN_IOS, 0);
	2643	+ pools->bs = bioset_create(pool_size, 0);
2644	2644	if (!pools->bs)
2645	2645	goto free_tio_pool_and_out;
2646	2646
	2647	+ if (integrity && bioset_integrity_create(pools->bs, pool_size))
	2648	+ goto free_bioset_and_out;
	2649	+
2647	2650	return pools;
	2651	+
	2652	+free_bioset_and_out:
	2653	+ bioset_free(pools->bs);
2648	2654
2649	2655	free_tio_pool_and_out:
2650	2656	mempool_destroy(pools->tio_pool);
...	...	@@ -149,7 +149,7 @@
149	149	/*
150	150	* Mempool operations
151	151	*/
152		-struct dm_md_mempools *dm_alloc_md_mempools(unsigned type);
	152	+struct dm_md_mempools *dm_alloc_md_mempools(unsigned type, unsigned integrity);
153	153	void dm_free_md_mempools(struct dm_md_mempools *pools);
154	154
155	155	#endif
...	...	@@ -210,8 +210,7 @@
210	210	blk_queue_merge_bvec(mddev->queue, linear_mergeable_bvec);
211	211	mddev->queue->backing_dev_info.congested_fn = linear_congested;
212	212	mddev->queue->backing_dev_info.congested_data = mddev;
213		- md_integrity_register(mddev);
214		- return 0;
	213	+ return md_integrity_register(mddev);
215	214	}
216	215
217	216	static void free_conf(struct rcu_head *head)
...	...	@@ -1803,8 +1803,12 @@
1803	1803	mdname(mddev));
1804	1804	return -EINVAL;
1805	1805	}
1806		- printk(KERN_NOTICE "md: data integrity on %s enabled\n",
1807		- mdname(mddev));
	1806	+ printk(KERN_NOTICE "md: data integrity enabled on %s\n", mdname(mddev));
	1807	+ if (bioset_integrity_create(mddev->bio_set, BIO_POOL_SIZE)) {
	1808	+ printk(KERN_ERR "md: failed to create integrity pool for %s\n",
	1809	+ mdname(mddev));
	1810	+ return -EINVAL;
	1811	+ }
1808	1812	return 0;
1809	1813	}
1810	1814	EXPORT_SYMBOL(md_integrity_register);
...	...	@@ -315,7 +315,7 @@
315	315	p->rdev = rdev;
316	316	goto abort;
317	317	}
318		- md_integrity_register(mddev);
	318	+ err = md_integrity_register(mddev);
319	319	}
320	320	abort:
321	321
...	...	@@ -489,7 +489,10 @@
489	489
490	490	mddev->queue->backing_dev_info.congested_fn = multipath_congested;
491	491	mddev->queue->backing_dev_info.congested_data = mddev;
492		- md_integrity_register(mddev);
	492	+
	493	+ if (md_integrity_register(mddev))
	494	+ goto out_free_conf;
	495	+
493	496	return 0;
494	497
495	498	out_free_conf:
...	...	@@ -379,8 +379,7 @@
379	379
380	380	blk_queue_merge_bvec(mddev->queue, raid0_mergeable_bvec);
381	381	dump_zones(mddev);
382		- md_integrity_register(mddev);
383		- return 0;
	382	+ return md_integrity_register(mddev);
384	383	}
385	384
386	385	static int raid0_stop(mddev_t *mddev)
...	...	@@ -1132,7 +1132,7 @@
1132	1132	p->rdev = rdev;
1133	1133	goto abort;
1134	1134	}
1135		- md_integrity_register(mddev);
	1135	+ err = md_integrity_register(mddev);
1136	1136	}
1137	1137	abort:
1138	1138
...	...	@@ -2017,8 +2017,7 @@
2017	2017
2018	2018	mddev->queue->backing_dev_info.congested_fn = raid1_congested;
2019	2019	mddev->queue->backing_dev_info.congested_data = mddev;
2020		- md_integrity_register(mddev);
2021		- return 0;
	2020	+ return md_integrity_register(mddev);
2022	2021	}
2023	2022
2024	2023	static int stop(mddev_t *mddev)
...	...	@@ -1188,7 +1188,7 @@
1188	1188	p->rdev = rdev;
1189	1189	goto abort;
1190	1190	}
1191		- md_integrity_register(mddev);
	1191	+ err = md_integrity_register(mddev);
1192	1192	}
1193	1193	abort:
1194	1194
...	...	@@ -2343,7 +2343,10 @@
2343	2343
2344	2344	if (conf->near_copies < conf->raid_disks)
2345	2345	blk_queue_merge_bvec(mddev->queue, raid10_mergeable_bvec);
2346		- md_integrity_register(mddev);
	2346	+
	2347	+ if (md_integrity_register(mddev))
	2348	+ goto out_free_conf;
	2349	+
2347	2350	return 0;
2348	2351
2349	2352	out_free_conf:
...	...	@@ -761,6 +761,9 @@
761	761	{
762	762	unsigned int max_slab = vecs_to_idx(BIO_MAX_PAGES);
763	763
	764	+ if (bs->bio_integrity_pool)
	765	+ return 0;
	766	+
764	767	bs->bio_integrity_pool =
765	768	mempool_create_slab_pool(pool_size, bip_slab[max_slab].slab);
766	769