[PATCH] block: propagate in_flight to whole disk on partition I/O

Tang Yizhou posted 1 patch 2 days, 6 hours ago
There is a newer version of this series
block/blk-core.c | 4 ++++
block/blk-mq.c   | 6 ++++++
2 files changed, 10 insertions(+)
[PATCH] block: propagate in_flight to whole disk on partition I/O
Posted by Tang Yizhou 2 days, 6 hours ago
From: Tang Yizhou <yizhou.tang@shopee.com>

Now when I/O is submitted to a partition, the per-CPU in_flight[]
counter is incremented only on the partition's block_device, not on the
underlying whole disk. This leads to a problem which can be shown by a
fio test:

lsblk
  NAME     MAJ:MIN RM  SIZE RO TYPE MOUNTPOINTS
  mydev    252:1    0   20G  0 disk
  └─mydev1 259:0    0   10G  0 part

iostat -xp 1
  Device       r/s        rkB/s      ... aqu-sz   %util
  mydev    128153.00  512612.00      ...  13.22   72.20
  mydev1   128154.00  512616.00      ...  13.22  100.00

%util is different between mydev and mydev1, which is unexpected.

This is the cumulative effect of a series of patches.  The key step is
commit 10ec5e86f9b8 ("block: merge part_{inc,dev}_in_flight into their
only callers"), which folded the whole-disk in_flight accounting into
generic_start_io_acct() and generic_end_io_acct().  Those two helpers
were then removed by commit e722fff238bb ("block: remove
generic_{start,end}_io_acct"), and from that point on the whole disk's
in_flight is no longer accounted at all.

Fix it by restoring the whole-disk in_flight accounting.

Fixes: e722fff238bb ("block: remove generic_{start,end}_io_acct")

Suggested-by: Leon Hwang <leon.huangfu@shopee.com>
Signed-off-by: Tang Yizhou <yizhou.tang@shopee.com>
---
 block/blk-core.c | 4 ++++
 block/blk-mq.c   | 6 ++++++
 2 files changed, 10 insertions(+)

diff --git a/block/blk-core.c b/block/blk-core.c
index 17450058ea6d..03f4b7015e69 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -1043,6 +1043,8 @@ unsigned long bdev_start_io_acct(struct block_device *bdev, enum req_op op,
 	part_stat_lock();
 	update_io_ticks(bdev, start_time, false);
 	part_stat_local_inc(bdev, in_flight[op_is_write(op)]);
+	if (bdev_is_partition(bdev))
+		part_stat_local_inc(bdev_whole(bdev), in_flight[op_is_write(op)]);
 	part_stat_unlock();
 
 	return start_time;
@@ -1074,6 +1076,8 @@ void bdev_end_io_acct(struct block_device *bdev, enum req_op op,
 	part_stat_add(bdev, sectors[sgrp], sectors);
 	part_stat_add(bdev, nsecs[sgrp], jiffies_to_nsecs(duration));
 	part_stat_local_dec(bdev, in_flight[op_is_write(op)]);
+	if (bdev_is_partition(bdev))
+		part_stat_local_dec(bdev_whole(bdev), in_flight[op_is_write(op)]);
 	part_stat_unlock();
 }
 EXPORT_SYMBOL(bdev_end_io_acct);
diff --git a/block/blk-mq.c b/block/blk-mq.c
index d0c37daf568f..60ead16f1496 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -1084,6 +1084,9 @@ static inline void blk_account_io_done(struct request *req, u64 now)
 		part_stat_add(req->part, nsecs[sgrp], now - req->start_time_ns);
 		part_stat_local_dec(req->part,
 				    in_flight[op_is_write(req_op(req))]);
+		if (bdev_is_partition(req->part))
+			part_stat_local_dec(bdev_whole(req->part),
+					    in_flight[op_is_write(req_op(req))]);
 		part_stat_unlock();
 	}
 }
@@ -1144,6 +1147,9 @@ static inline void blk_account_io_start(struct request *req)
 	part_stat_lock();
 	update_io_ticks(req->part, jiffies, false);
 	part_stat_local_inc(req->part, in_flight[op_is_write(req_op(req))]);
+	if (bdev_is_partition(req->part))
+		part_stat_local_inc(bdev_whole(req->part),
+				    in_flight[op_is_write(req_op(req))]);
 	part_stat_unlock();
 }
 
-- 
2.43.0

Re: [PATCH] block: propagate in_flight to whole disk on partition I/O
Posted by Christoph Hellwig 2 days, 5 hours ago
On Fri, May 22, 2026 at 07:37:51PM +0800, Tang Yizhou wrote:
> --- a/block/blk-core.c
> +++ b/block/blk-core.c
> @@ -1043,6 +1043,8 @@ unsigned long bdev_start_io_acct(struct block_device *bdev, enum req_op op,
>  	part_stat_lock();
>  	update_io_ticks(bdev, start_time, false);
>  	part_stat_local_inc(bdev, in_flight[op_is_write(op)]);
> +	if (bdev_is_partition(bdev))
> +		part_stat_local_inc(bdev_whole(bdev), in_flight[op_is_write(op)]);

overly lone line.

> +	if (bdev_is_partition(bdev))
> +		part_stat_local_dec(bdev_whole(bdev), in_flight[op_is_write(op)]);

Same.

>  }
> @@ -1144,6 +1147,9 @@ static inline void blk_account_io_start(struct request *req)
>  	part_stat_lock();
>  	update_io_ticks(req->part, jiffies, false);
>  	part_stat_local_inc(req->part, in_flight[op_is_write(req_op(req))]);
> +	if (bdev_is_partition(req->part))
> +		part_stat_local_inc(bdev_whole(req->part),
> +				    in_flight[op_is_write(req_op(req))]);

and tis duplicates the above logic.  Mabye factor the common code
into two little helpers?
Re: [PATCH] block: propagate in_flight to whole disk on partition I/O
Posted by Tang Yizhou 2 days, 5 hours ago
On Fri, May 22, 2026 at 8:12 PM Christoph Hellwig <hch@lst.de> wrote:
>
> On Fri, May 22, 2026 at 07:37:51PM +0800, Tang Yizhou wrote:
> > --- a/block/blk-core.c
> > +++ b/block/blk-core.c
> > @@ -1043,6 +1043,8 @@ unsigned long bdev_start_io_acct(struct block_device *bdev, enum req_op op,
> >       part_stat_lock();
> >       update_io_ticks(bdev, start_time, false);
> >       part_stat_local_inc(bdev, in_flight[op_is_write(op)]);
> > +     if (bdev_is_partition(bdev))
> > +             part_stat_local_inc(bdev_whole(bdev), in_flight[op_is_write(op)]);
>
> overly lone line.

OK. I will update in the next patch.

>
> > +     if (bdev_is_partition(bdev))
> > +             part_stat_local_dec(bdev_whole(bdev), in_flight[op_is_write(op)]);
>
> Same.
>
> >  }
> > @@ -1144,6 +1147,9 @@ static inline void blk_account_io_start(struct request *req)
> >       part_stat_lock();
> >       update_io_ticks(req->part, jiffies, false);
> >       part_stat_local_inc(req->part, in_flight[op_is_write(req_op(req))]);
> > +     if (bdev_is_partition(req->part))
> > +             part_stat_local_inc(bdev_whole(req->part),
> > +                                 in_flight[op_is_write(req_op(req))]);
>
> and tis duplicates the above logic.  Mabye factor the common code
> into two little helpers?

Sure.

>