Skip to contentSkip to conversation

fix and cleanup device mapper bioset initialization

PATCH New Christoph Hellwig 2 participants 8 messages Jun 8, 2022 – Jun 9, 2022

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

Hi Mike,

the first patch fixes the device mapper bioset to restore the previous behavior of preallocating biosets instead of allocating them at bind time, and to actually allocate pools for the integrity data. The others are cleanups on top of that.

Diffstat:

 block/bio.c           |   20 -------
 drivers/md/dm-core.h  |   14 ++++-
 drivers/md/dm-rq.c    |    3 -
 drivers/md/dm-table.c |   66 ++++++++++++++----------
 drivers/md/dm.c       |  136 +++++++++-----------------------------------------
 drivers/md/dm.h       |    5 -
 include/linux/bio.h   |    1 
 7 files changed, 78 insertions(+), 167 deletions(-)
#1 lore raw

PATCH 1/4 dm: fix bio_set allocation

The use of bioset_init_from_src mean that the pre-allocated pools weren't used for anything except parameter passing, and the integrity pool creation got completely lost for the actual live mapped_device. Fix that by assigning the actual preallocated dm_md_mempools to the mapped_device and using that for I/O instead of creating new mempools.

Fixes 2a2a4c510b76 ("dm: use bioset_init_from_src() to copy bio_set")
---
 drivers/md/dm-core.h  | 11 ++++--
 drivers/md/dm-rq.c    |  2 +-
 drivers/md/dm-table.c | 11 ------
 drivers/md/dm.c       | 84 +++++++++++++------------------------------
 drivers/md/dm.h       |  2 --
 5 files changed, 35 insertions(+), 75 deletions(-)
drivers/md/dm-core.h +9 -2
index d21648a923ea9..54c0473a51dde 100644
@@ -33,6 +33,14 @@ struct dm_kobject_holder {
  * access their members!
  */
 
+/*
+ * For mempools pre-allocation at the table loading time.
+ */
+struct dm_md_mempools {
+	struct bio_set bs;
+	struct bio_set io_bs;
+};
+
 struct mapped_device {
 	struct mutex suspend_lock;
 
@@ -110,8 +118,7 @@ struct mapped_device {
 	/*
 	 * io objects are allocated from here.
 	 */
-	struct bio_set io_bs;
-	struct bio_set bs;
+	struct dm_md_mempools *mempools;
 
 	/* kobject and completion */
 	struct dm_kobject_holder kobj_holder;
drivers/md/dm-rq.c +1 -1
index 6087cdcaad46d..a83b98a8d2a99 100644
@@ -319,7 +319,7 @@ static int setup_clone(struct request *clone, struct request *rq,
 {
 	int r;
 
-	r = blk_rq_prep_clone(clone, rq, &tio->md->bs, gfp_mask,
+	r = blk_rq_prep_clone(clone, rq, &tio->md->mempools->bs, gfp_mask,
 			      dm_rq_bio_constructor, tio);
 	if (r)
 		return r;
drivers/md/dm-table.c +0 -11
index 0e833a154b31d..bd539afbfe88f 100644
@@ -1038,17 +1038,6 @@ static int dm_table_alloc_md_mempools(struct dm_table *t, struct mapped_device *
 	return 0;
 }
 
-void dm_table_free_md_mempools(struct dm_table *t)
-{
-	dm_free_md_mempools(t->mempools);
-	t->mempools = NULL;
-}
-
-struct dm_md_mempools *dm_table_get_md_mempools(struct dm_table *t)
-{
-	return t->mempools;
-}
-
 static int setup_indexes(struct dm_table *t)
 {
 	int i;
drivers/md/dm.c +25 -59

121 lines not on this page · raw message

drivers/md/dm.h +0 -2
index 3f89664fea010..a8405ce305a96 100644
@@ -71,8 +71,6 @@ struct dm_target *dm_table_get_immutable_target(struct dm_table *t);
 struct dm_target *dm_table_get_wildcard_target(struct dm_table *t);
 bool dm_table_bio_based(struct dm_table *t);
 bool dm_table_request_based(struct dm_table *t);
-void dm_table_free_md_mempools(struct dm_table *t);
-struct dm_md_mempools *dm_table_get_md_mempools(struct dm_table *t);
 
 void dm_lock_md_type(struct mapped_device *md);
 void dm_unlock_md_type(struct mapped_device *md);
Signature · 1 line
2.30.2
#2 lore raw

PATCH 2/4 block: remove bioset_init_from_src

In reply to Christoph Hellwig

Unused now, and the interface never really made a whole lot of sense to start with.

---
 block/bio.c         | 20 --------------------
 include/linux/bio.h |  1 -
 2 files changed, 21 deletions(-)
block/bio.c +0 -20
index be3937b84e68a..f3cce7f267a98 100644
@@ -1745,26 +1745,6 @@ int bioset_init(struct bio_set *bs,
 }
 EXPORT_SYMBOL(bioset_init);
 
-/*
- * Initialize and setup a new bio_set, based on the settings from
- * another bio_set.
- */
-int bioset_init_from_src(struct bio_set *bs, struct bio_set *src)
-{
-	int flags;
-
-	flags = 0;
-	if (src->bvec_pool.min_nr)
-		flags |= BIOSET_NEED_BVECS;
-	if (src->rescue_workqueue)
-		flags |= BIOSET_NEED_RESCUER;
-	if (src->cache)
-		flags |= BIOSET_PERCPU_CACHE;
-
-	return bioset_init(bs, src->bio_pool.min_nr, src->front_pad, flags);
-}
-EXPORT_SYMBOL(bioset_init_from_src);
-
 static int __init init_bio(void)
 {
 	int i;
include/linux/bio.h +0 -1
index 1cf3738ef1ea6..992ee987f2738 100644
@@ -403,7 +403,6 @@ enum {
 extern int bioset_init(struct bio_set *, unsigned int, unsigned int, int flags);
 extern void bioset_exit(struct bio_set *);
 extern int biovec_init_pool(mempool_t *pool, int pool_entries);
-extern int bioset_init_from_src(struct bio_set *bs, struct bio_set *src);
 
 struct bio *bio_alloc_bioset(struct block_device *bdev, unsigned short nr_vecs,
 			     unsigned int opf, gfp_t gfp_mask,
Signature · 1 line
2.30.2
#3 lore raw

PATCH 3/4 dm: unexport dm_get_reserved_rq_based_ios

In reply to Christoph Hellwig

dm_get_reserved_rq_based_ios is only used in the core dm code, so remove the export.

---
 drivers/md/dm-rq.c | 1 -
 1 file changed, 1 deletion(-)
drivers/md/dm-rq.c +0 -1
index a83b98a8d2a99..4f49bbcce4f1a 100644
@@ -43,7 +43,6 @@ unsigned dm_get_reserved_rq_based_ios(void)
 	return __dm_get_module_param(&reserved_rq_based_ios,
 				     RESERVED_REQUEST_BASED_IOS, DM_RESERVED_MAX_IOS);
 }
-EXPORT_SYMBOL_GPL(dm_get_reserved_rq_based_ios);
 
 static unsigned dm_get_blk_mq_nr_hw_queues(void)
 {
Signature · 1 line
2.30.2
#4 lore raw

PATCH 4/4 dm: refactor dm_md_mempool allocation

In reply to Christoph Hellwig

The current split between dm_table_alloc_md_mempools and dm_alloc_md_mempools is rather arbitrary, so merge the two into one easy to follow function.

---
 drivers/md/dm-core.h  |  3 +++
 drivers/md/dm-table.c | 57 +++++++++++++++++++++++++++++++------------
 drivers/md/dm.c       | 52 ---------------------------------------
 drivers/md/dm.h       |  3 ---
 4 files changed, 44 insertions(+), 71 deletions(-)
drivers/md/dm-core.h +3 -0
index 54c0473a51dde..eea3922f1abc6 100644
@@ -230,6 +230,9 @@ struct dm_target_io {
 	sector_t old_sector;
 	struct bio clone;
 };
+#define DM_TARGET_IO_BIO_OFFSET (offsetof(struct dm_target_io, clone))
+#define DM_IO_BIO_OFFSET \
+	(offsetof(struct dm_target_io, clone) + offsetof(struct dm_io, tio))
 
 /*
  * dm_target_io flags
drivers/md/dm-table.c +41 -16

79 lines not on this page · raw message

drivers/md/dm.c +0 -52

64 lines not on this page · raw message

drivers/md/dm.h +0 -3
index a8405ce305a96..62816b647f827 100644
@@ -218,9 +218,6 @@ void dm_kcopyd_exit(void);
 /*
  * Mempool operations
  */
-struct dm_md_mempools *dm_alloc_md_mempools(struct mapped_device *md, enum dm_queue_mode type,
-					    unsigned per_io_data_size, unsigned min_pool_size,
-					    bool integrity, bool poll);
 void dm_free_md_mempools(struct dm_md_mempools *pools);
 
 /*
Signature · 1 line
2.30.2
maintainer
#5 lore raw
In reply to Christoph Hellwig
Quoted 8 lines · Christoph Hellwig

Hi Mike,

the first patch fixes the device mapper bioset to restore the previous behavior of preallocating biosets instead of allocating them at bind time, and to actually allocate pools for the integrity data. The others are cleanups on top of that.

Diffstat:

Quoted 9 lines
 block/bio.c           |   20 -------
 drivers/md/dm-core.h  |   14 ++++-
 drivers/md/dm-rq.c    |    3 -
 drivers/md/dm-table.c |   66 ++++++++++++++----------
 drivers/md/dm.c       |  136 +++++++++-----------------------------------------
 drivers/md/dm.h       |    5 -
 include/linux/bio.h   |    1 
 7 files changed, 78 insertions(+), 167 deletions(-)

All looks good to me. Are you OK with me picking up the first 3 to send to Linus for 5.19-rc2 (given the integrity bioset fix)?

And hold patch 4 until 5.20 merge? Or would you prefer that cleanup to land now too?

Mike

--
dm-devel mailing list
[email protected]
https://listman.redhat.com/mailman/listinfo/dm-devel

#6 lore raw
Quoted 4 lines · Mike Snitzer

All looks good to me. Are you OK with me picking up the first 3 to send to Linus for 5.19-rc2 (given the integrity bioset fix)?

And hold patch 4 until 5.20 merge?

Sounds good to me.

Quoted 1 line

Or would you prefer that cleanup to land now too?

I don't think Linus would like that :) In fact even patch 3 might be 5.20 material.

maintainer
#7 lore raw
Quoted 4 lines · Christoph Hellwig
Quoted 3 lines · Mike Snitzer

All looks good to me. Are you OK with me picking up the first 3 to send to Linus for 5.19-rc2 (given the integrity bioset fix)?

Quoted 8 lines
Quoted 1 line

And hold patch 4 until 5.20 merge?

Sounds good to me.

Quoted 1 line

Or would you prefer that cleanup to land now too?

I don't think Linus would like that :) In fact even patch 3 might be 5.20 material.

Ha, yeah I agree. I'll just send the first 2 (likely tomorrow after more testing today).

Thanks.