[PATCH v2 0/3] block: improve iostat for md/bcache partitions

[PATCH v2 0/3] block: improve iostat for md/bcache partitions

[Song Liu]

Currently, devices like md, bcache uses disk_[start|end]_io_acct to report
iostat. These functions couldn't get proper iostat for partitions on these
devices.

This set resolves this issue by introducing part_[begin|end]_io_acct, and
using them in md and bcache code.

Changes v1 => v2:
1. Refactor the code, as suggested by Christoph.
2. Include Coly's Reviewed-by tag.

Song Liu (3):
  block: introduce part_[begin|end]_io_acct
  md: use part_[begin|end]_io_acct instead of disk_[begin|end]_io_acct
  bcache: use part_[begin|end]_io_acct instead of
    disk_[begin|end]_io_acct

 block/blk-core.c            | 39 +++++++++++++++++++++++++++++++------
 drivers/md/bcache/request.c | 10 ++++++----
 drivers/md/md.c             |  8 ++++----
 include/linux/blkdev.h      |  5 +++++
 4 files changed, 48 insertions(+), 14 deletions(-)

--
2.24.1

[PATCH v2 1/3] block: introduce part_[begin|end]_io_acct

[Song Liu]

These functions can be used to enable iostat for partitions on devices
like md, bcache.

Signed-off-by: Song Liu <songliubraving@fb.com>
---
 block/blk-core.c       | 39 +++++++++++++++++++++++++++++++++------
 include/linux/blkdev.h |  5 +++++
 2 files changed, 38 insertions(+), 6 deletions(-)

diff --git a/block/blk-core.c b/block/blk-core.c
index 93104c7470e8a..02953e9db3dd0 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -1455,10 +1455,9 @@ void blk_account_io_start(struct request *rq)
 	part_stat_unlock();
 }
 
-unsigned long disk_start_io_acct(struct gendisk *disk, unsigned int sectors,
-		unsigned int op)
+static unsigned long __part_start_io_acct(struct hd_struct *part,
+					  unsigned int sectors, unsigned int op)
 {
-	struct hd_struct *part = &disk->part0;
 	const int sgrp = op_stat_group(op);
 	unsigned long now = READ_ONCE(jiffies);
 
@@ -1471,12 +1470,26 @@ unsigned long disk_start_io_acct(struct gendisk *disk, unsigned int sectors,
 
 	return now;
 }
+
+unsigned long part_start_io_acct(struct gendisk *disk, struct hd_struct **part,
+				 struct bio *bio)
+{
+	*part = disk_map_sector_rcu(disk, bio->bi_iter.bi_sector);
+
+	return __part_start_io_acct(*part, bio_sectors(bio), bio_op(bio));
+}
+EXPORT_SYMBOL(part_start_io_acct);
+
+unsigned long disk_start_io_acct(struct gendisk *disk, unsigned int sectors,
+				 unsigned int op)
+{
+	return __part_start_io_acct(&disk->part0, sectors, op);
+}
 EXPORT_SYMBOL(disk_start_io_acct);
 
-void disk_end_io_acct(struct gendisk *disk, unsigned int op,
-		unsigned long start_time)
+static void __part_end_io_acct(struct hd_struct *part, unsigned int op,
+			       unsigned long start_time)
 {
-	struct hd_struct *part = &disk->part0;
 	const int sgrp = op_stat_group(op);
 	unsigned long now = READ_ONCE(jiffies);
 	unsigned long duration = now - start_time;
@@ -1487,6 +1500,20 @@ void disk_end_io_acct(struct gendisk *disk, unsigned int op,
 	part_stat_local_dec(part, in_flight[op_is_write(op)]);
 	part_stat_unlock();
 }
+
+void part_end_io_acct(struct hd_struct *part, struct bio *bio,
+		      unsigned long start_time)
+{
+	__part_end_io_acct(part, bio_op(bio), start_time);
+	hd_struct_put(part);
+}
+EXPORT_SYMBOL(part_end_io_acct);
+
+void disk_end_io_acct(struct gendisk *disk, unsigned int op,
+		      unsigned long start_time)
+{
+	__part_end_io_acct(&disk->part0, op, start_time);
+}
 EXPORT_SYMBOL(disk_end_io_acct);
 
 /*
diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index 285b59cfc0646..d2ee868b16a51 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -1931,6 +1931,11 @@ unsigned long disk_start_io_acct(struct gendisk *disk, unsigned int sectors,
 void disk_end_io_acct(struct gendisk *disk, unsigned int op,
 		unsigned long start_time);
 
+unsigned long part_start_io_acct(struct gendisk *disk, struct hd_struct **part,
+				 struct bio *bio);
+void part_end_io_acct(struct hd_struct *part, struct bio *bio,
+		      unsigned long start_time);
+
 /**
  * bio_start_io_acct - start I/O accounting for bio based drivers
  * @bio:	bio to start account for
-- 
2.24.1

Re: [PATCH v2 1/3] block: introduce part_[begin|end]_io_acct

[Christoph Hellwig]

On Tue, Aug 18, 2020 at 03:26:43PM -0700, Song Liu wrote:
> These functions can be used to enable iostat for partitions on devices
> like md, bcache.
> 
> Signed-off-by: Song Liu <songliubraving@fb.com>

EXPORT_SYMBOL_GPL for new block layer functionality, please.

Otherwise this looks good:

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH v2 1/3] block: introduce part_[begin|end]_io_acct

[Song Liu]

> On Aug 30, 2020, at 11:53 PM, Christoph Hellwig <hch@infradead.org> wrote:
> 
> On Tue, Aug 18, 2020 at 03:26:43PM -0700, Song Liu wrote:
>> These functions can be used to enable iostat for partitions on devices
>> like md, bcache.
>> 
>> Signed-off-by: Song Liu <songliubraving@fb.com>
> 
> EXPORT_SYMBOL_GPL for new block layer functionality, please.
> 
> Otherwise this looks good:
> 
> Reviewed-by: Christoph Hellwig <hch@lst.de>

Thanks Christoph! Will fix and resend v3. 

[PATCH v2 2/3] md: use part_[begin|end]_io_acct instead of disk_[begin|end]_io_acct

[Song Liu]

This enables proper statistics in /proc/diskstats for md partitions.

Signed-off-by: Song Liu <songliubraving@fb.com>
---
 drivers/md/md.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/md/md.c b/drivers/md/md.c
index 6b511c9007d38..5119370c09a6b 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -470,6 +470,7 @@ struct md_io {
 	bio_end_io_t *orig_bi_end_io;
 	void *orig_bi_private;
 	unsigned long start_time;
+	struct hd_struct *part;
 };
 
 static void md_end_io(struct bio *bio)
@@ -477,7 +478,7 @@ static void md_end_io(struct bio *bio)
 	struct md_io *md_io = bio->bi_private;
 	struct mddev *mddev = md_io->mddev;
 
-	disk_end_io_acct(mddev->gendisk, bio_op(bio), md_io->start_time);
+	part_end_io_acct(md_io->part, bio, md_io->start_time);
 
 	bio->bi_end_io = md_io->orig_bi_end_io;
 	bio->bi_private = md_io->orig_bi_private;
@@ -523,9 +524,8 @@ static blk_qc_t md_submit_bio(struct bio *bio)
 		bio->bi_end_io = md_end_io;
 		bio->bi_private = md_io;
 
-		md_io->start_time = disk_start_io_acct(mddev->gendisk,
-						       bio_sectors(bio),
-						       bio_op(bio));
+		md_io->start_time = part_start_io_acct(mddev->gendisk,
+						       &md_io->part, bio);
 	}
 
 	/* bio could be mergeable after passing to underlayer */
-- 
2.24.1

Re: [PATCH v2 2/3] md: use part_[begin|end]_io_acct instead of disk_[begin|end]_io_acct

[Christoph Hellwig]

On Tue, Aug 18, 2020 at 03:26:44PM -0700, Song Liu wrote:
> This enables proper statistics in /proc/diskstats for md partitions.
> 
> Signed-off-by: Song Liu <songliubraving@fb.com>

Looks good,

Reviewed-by: Christoph Hellwig <hch@lst.de>

[PATCH v2 3/3] bcache: use part_[begin|end]_io_acct instead of disk_[begin|end]_io_acct

[Song Liu]

This enables proper statistics in /proc/diskstats for bcache partitions.

Reviewed-by: Coly Li <colyli@suse.de>
Signed-off-by: Song Liu <songliubraving@fb.com>
---
 drivers/md/bcache/request.c | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

diff --git a/drivers/md/bcache/request.c b/drivers/md/bcache/request.c
index 7de82c67597ce..5a0b6eef33748 100644
--- a/drivers/md/bcache/request.c
+++ b/drivers/md/bcache/request.c
@@ -475,6 +475,7 @@ struct search {
 	unsigned int		read_dirty_data:1;
 	unsigned int		cache_missed:1;
 
+	struct hd_struct	*part;
 	unsigned long		start_time;
 
 	struct btree_op		op;
@@ -669,7 +670,7 @@ static void bio_complete(struct search *s)
 {
 	if (s->orig_bio) {
 		/* Count on bcache device */
-		disk_end_io_acct(s->d->disk, bio_op(s->orig_bio), s->start_time);
+		part_end_io_acct(s->part, s->orig_bio, s->start_time);
 
 		trace_bcache_request_end(s->d, s->orig_bio);
 		s->orig_bio->bi_status = s->iop.status;
@@ -731,7 +732,7 @@ static inline struct search *search_alloc(struct bio *bio,
 	s->write		= op_is_write(bio_op(bio));
 	s->read_dirty_data	= 0;
 	/* Count on the bcache device */
-	s->start_time		= disk_start_io_acct(d->disk, bio_sectors(bio), bio_op(bio));
+	s->start_time		= part_start_io_acct(d->disk, &s->part, bio);
 	s->iop.c		= d->c;
 	s->iop.bio		= NULL;
 	s->iop.inode		= d->id;
@@ -1072,6 +1073,7 @@ struct detached_dev_io_private {
 	unsigned long		start_time;
 	bio_end_io_t		*bi_end_io;
 	void			*bi_private;
+	struct hd_struct	*part;
 };
 
 static void detached_dev_end_io(struct bio *bio)
@@ -1083,7 +1085,7 @@ static void detached_dev_end_io(struct bio *bio)
 	bio->bi_private = ddip->bi_private;
 
 	/* Count on the bcache device */
-	disk_end_io_acct(ddip->d->disk, bio_op(bio), ddip->start_time);
+	part_end_io_acct(ddip->part, bio, ddip->start_time);
 
 	if (bio->bi_status) {
 		struct cached_dev *dc = container_of(ddip->d,
@@ -1109,7 +1111,7 @@ static void detached_dev_do_request(struct bcache_device *d, struct bio *bio)
 	ddip = kzalloc(sizeof(struct detached_dev_io_private), GFP_NOIO);
 	ddip->d = d;
 	/* Count on the bcache device */
-	ddip->start_time = disk_start_io_acct(d->disk, bio_sectors(bio), bio_op(bio));
+	ddip->start_time = part_start_io_acct(d->disk, &ddip->part, bio);
 	ddip->bi_end_io = bio->bi_end_io;
 	ddip->bi_private = bio->bi_private;
 	bio->bi_end_io = detached_dev_end_io;
-- 
2.24.1

Re: [PATCH v2 3/3] bcache: use part_[begin|end]_io_acct instead of disk_[begin|end]_io_acct

[Christoph Hellwig]

On Tue, Aug 18, 2020 at 03:26:45PM -0700, Song Liu wrote:
> This enables proper statistics in /proc/diskstats for bcache partitions.
> 
> Reviewed-by: Coly Li <colyli@suse.de>
> Signed-off-by: Song Liu <songliubraving@fb.com>

Looks good,

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH v2 0/3] block: improve iostat for md/bcache partitions

[Song Liu]

> On Aug 18, 2020, at 3:26 PM, Song Liu <songliubraving@fb.com> wrote:
> 
> Currently, devices like md, bcache uses disk_[start|end]_io_acct to report
> iostat. These functions couldn't get proper iostat for partitions on these
> devices.
> 
> This set resolves this issue by introducing part_[begin|end]_io_acct, and
> using them in md and bcache code.
> 
> Changes v1 => v2:
> 1. Refactor the code, as suggested by Christoph.
> 2. Include Coly's Reviewed-by tag.
> 
> Song Liu (3):
>  block: introduce part_[begin|end]_io_acct
>  md: use part_[begin|end]_io_acct instead of disk_[begin|end]_io_acct
>  bcache: use part_[begin|end]_io_acct instead of
>    disk_[begin|end]_io_acct
> 
> block/blk-core.c            | 39 +++++++++++++++++++++++++++++++------
> drivers/md/bcache/request.c | 10 ++++++----
> drivers/md/md.c             |  8 ++++----
> include/linux/blkdev.h      |  5 +++++
> 4 files changed, 48 insertions(+), 14 deletions(-)

Hi Christoph, 

Does this version look good to you?

Thanks,
Song