Discussion:
[dm-devel] [PATCH 2/3] block: switch to per-cpu in-flight counters
Mikulas Patocka
2018-11-28 00:42:13 UTC
Permalink
Now when part_round_stats is gone, we can switch to per-cpu in-flight
counters.

We use the local-atomic type local_t, so that if part_inc_in_flight or
part_dec_in_flight is reentrantly called from an interrupt, the value will
be correct.

The other counters could be corrupted due to reentrant interrupt, but the
corruption only results in slight counter skew - the in_flight counter
must be exact, so it needs local_t.

Signed-off-by: Mikulas Patocka <***@redhat.com>

---
block/bio.c | 4 ++--
block/blk-core.c | 4 ++--
block/blk-merge.c | 2 +-
block/genhd.c | 47 +++++++++++++++++++++++++++++++++++------------
drivers/md/dm.c | 4 +---
include/linux/genhd.h | 7 ++++---
6 files changed, 45 insertions(+), 23 deletions(-)

Index: linux-block/block/genhd.c
===================================================================
--- linux-block.orig/block/genhd.c 2018-11-28 00:09:59.000000000 +0100
+++ linux-block/block/genhd.c 2018-11-28 00:09:59.000000000 +0100
@@ -45,53 +45,76 @@ static void disk_add_events(struct gendi
static void disk_del_events(struct gendisk *disk);
static void disk_release_events(struct gendisk *disk);

-void part_inc_in_flight(struct request_queue *q, struct hd_struct *part, int rw)
+void part_inc_in_flight(struct request_queue *q, int cpu, struct hd_struct *part, int rw)
{
if (queue_is_mq(q))
return;

- atomic_inc(&part->in_flight[rw]);
+ local_inc(per_cpu_ptr(part->dkstats, cpu)->in_flight);
if (part->partno)
- atomic_inc(&part_to_disk(part)->part0.in_flight[rw]);
+ local_inc(per_cpu_ptr(part_to_disk(part)->part0.dkstats, cpu)->in_flight);
}

-void part_dec_in_flight(struct request_queue *q, struct hd_struct *part, int rw)
+void part_dec_in_flight(struct request_queue *q, int cpu, struct hd_struct *part, int rw)
{
if (queue_is_mq(q))
return;

- atomic_dec(&part->in_flight[rw]);
+ local_dec(per_cpu_ptr(part->dkstats, cpu)->in_flight);
if (part->partno)
- atomic_dec(&part_to_disk(part)->part0.in_flight[rw]);
+ local_dec(per_cpu_ptr(part_to_disk(part)->part0.dkstats, cpu)->in_flight);
}

void part_in_flight(struct request_queue *q, struct hd_struct *part,
unsigned int inflight[2])
{
+ int cpu;
+
if (queue_is_mq(q)) {
blk_mq_in_flight(q, part, inflight);
return;
}

- inflight[0] = atomic_read(&part->in_flight[0]) +
- atomic_read(&part->in_flight[1]);
+ inflight[0] = 0;
+ for_each_possible_cpu(cpu) {
+ inflight[0] += local_read(&per_cpu_ptr(part->dkstats, cpu)->in_flight[0]) +
+ local_read(&per_cpu_ptr(part->dkstats, cpu)->in_flight[1]);
+ }
+ if ((int)inflight[0] < 0)
+ inflight[0] = 0;
+
if (part->partno) {
part = &part_to_disk(part)->part0;
- inflight[1] = atomic_read(&part->in_flight[0]) +
- atomic_read(&part->in_flight[1]);
+ inflight[1] = 0;
+ for_each_possible_cpu(cpu) {
+ inflight[1] += local_read(&per_cpu_ptr(part->dkstats, cpu)->in_flight[0]) +
+ local_read(&per_cpu_ptr(part->dkstats, cpu)->in_flight[1]);
+ }
+ if ((int)inflight[1] < 0)
+ inflight[1] = 0;
}
}

void part_in_flight_rw(struct request_queue *q, struct hd_struct *part,
unsigned int inflight[2])
{
+ int cpu;
+
if (queue_is_mq(q)) {
blk_mq_in_flight_rw(q, part, inflight);
return;
}

- inflight[0] = atomic_read(&part->in_flight[0]);
- inflight[1] = atomic_read(&part->in_flight[1]);
+ inflight[0] = 0;
+ inflight[1] = 0;
+ for_each_possible_cpu(cpu) {
+ inflight[0] += local_read(&per_cpu_ptr(part->dkstats, cpu)->in_flight[0]);
+ inflight[1] += local_read(&per_cpu_ptr(part->dkstats, cpu)->in_flight[1]);
+ }
+ if ((int)inflight[0] < 0)
+ inflight[0] = 0;
+ if ((int)inflight[1] < 0)
+ inflight[1] = 0;
}

struct hd_struct *__disk_get_part(struct gendisk *disk, int partno)
Index: linux-block/include/linux/genhd.h
===================================================================
--- linux-block.orig/include/linux/genhd.h 2018-11-28 00:09:59.000000000 +0100
+++ linux-block/include/linux/genhd.h 2018-11-28 00:09:59.000000000 +0100
@@ -17,6 +17,7 @@
#include <linux/percpu-refcount.h>
#include <linux/uuid.h>
#include <linux/blk_types.h>
+#include <asm/local.h>

#ifdef CONFIG_BLOCK

@@ -89,6 +90,7 @@ struct disk_stats {
unsigned long merges[NR_STAT_GROUPS];
unsigned long io_ticks;
unsigned long time_in_queue;
+ local_t in_flight[2];
};

#define PARTITION_META_INFO_VOLNAMELTH 64
@@ -122,7 +124,6 @@ struct hd_struct {
int make_it_fail;
#endif
unsigned long stamp;
- atomic_t in_flight[2];
#ifdef CONFIG_SMP
struct disk_stats __percpu *dkstats;
#else
@@ -380,9 +381,9 @@ void part_in_flight(struct request_queue
unsigned int inflight[2]);
void part_in_flight_rw(struct request_queue *q, struct hd_struct *part,
unsigned int inflight[2]);
-void part_dec_in_flight(struct request_queue *q, struct hd_struct *part,
+void part_dec_in_flight(struct request_queue *q, int cpu, struct hd_struct *part,
int rw);
-void part_inc_in_flight(struct request_queue *q, struct hd_struct *part,
+void part_inc_in_flight(struct request_queue *q, int cpu, struct hd_struct *part,
int rw);

static inline struct partition_meta_info *alloc_part_info(struct gendisk *disk)
Index: linux-block/block/bio.c
===================================================================
--- linux-block.orig/block/bio.c 2018-11-28 00:09:59.000000000 +0100
+++ linux-block/block/bio.c 2018-11-28 00:09:59.000000000 +0100
@@ -1688,7 +1688,7 @@ void generic_start_io_acct(struct reques
update_io_ticks(cpu, part, jiffies);
part_stat_inc(cpu, part, ios[sgrp]);
part_stat_add(cpu, part, sectors[sgrp], sectors);
- part_inc_in_flight(q, part, op_is_write(op));
+ part_inc_in_flight(q, cpu, part, op_is_write(op));

part_stat_unlock();
}
@@ -1707,7 +1707,7 @@ void generic_end_io_acct(struct request_
part_stat_add(cpu, part, time_in_queue, duration);
if (part->partno)
part_stat_add(cpu, &part_to_disk(part)->part0, time_in_queue, duration);
- part_dec_in_flight(q, part, op_is_write(req_op));
+ part_dec_in_flight(q, cpu, part, op_is_write(req_op));

part_stat_unlock();
}
Index: linux-block/block/blk-merge.c
===================================================================
--- linux-block.orig/block/blk-merge.c 2018-11-28 00:09:59.000000000 +0100
+++ linux-block/block/blk-merge.c 2018-11-28 00:09:59.000000000 +0100
@@ -690,7 +690,7 @@ static void blk_account_io_merge(struct
cpu = part_stat_lock();
part = req->part;

- part_dec_in_flight(req->q, part, rq_data_dir(req));
+ part_dec_in_flight(req->q, cpu, part, rq_data_dir(req));

hd_struct_put(part);
part_stat_unlock();
Index: linux-block/block/blk-core.c
===================================================================
--- linux-block.orig/block/blk-core.c 2018-11-28 00:09:59.000000000 +0100
+++ linux-block/block/blk-core.c 2018-11-28 00:09:59.000000000 +0100
@@ -1356,7 +1356,7 @@ void blk_account_io_done(struct request
part_stat_add(cpu, part, nsecs[sgrp], now - req->start_time_ns);
part_stat_add(cpu, part, time_in_queue, nsecs_to_jiffies64(now - req->start_time_ns));
part_stat_add(cpu, &part_to_disk(part)->part0, time_in_queue, nsecs_to_jiffies64(now - req->start_time_ns));
- part_dec_in_flight(req->q, part, rq_data_dir(req));
+ part_dec_in_flight(req->q, cpu, part, rq_data_dir(req));

hd_struct_put(part);
part_stat_unlock();
@@ -1391,7 +1391,7 @@ void blk_account_io_start(struct request
part = &rq->rq_disk->part0;
hd_struct_get(part);
}
- part_inc_in_flight(rq->q, part, rw);
+ part_inc_in_flight(rq->q, cpu, part, rw);
rq->part = part;
}

Index: linux-block/drivers/md/dm.c
===================================================================
--- linux-block.orig/drivers/md/dm.c 2018-11-28 00:09:59.000000000 +0100
+++ linux-block/drivers/md/dm.c 2018-11-28 00:09:59.000000000 +0100
@@ -663,8 +663,7 @@ static void start_io_acct(struct dm_io *
generic_start_io_acct(md->queue, bio_op(bio), bio_sectors(bio),
&dm_disk(md)->part0);

- atomic_set(&dm_disk(md)->part0.in_flight[rw],
- atomic_inc_return(&md->pending[rw]));
+ atomic_inc(&md->pending[rw]);

if (unlikely(dm_stats_used(&md->stats)))
dm_stats_account_io(&md->stats, bio_data_dir(bio),
@@ -693,7 +692,6 @@ static void end_io_acct(struct dm_io *io
* a flush.
*/
pending = atomic_dec_return(&md->pending[rw]);
- atomic_set(&dm_disk(md)->part0.in_flight[rw], pending);
pending += atomic_read(&md->pending[rw^0x1]);

/* nudge anyone waiting on suspend queue */
Mike Snitzer
2018-11-29 21:57:47 UTC
Permalink
On Tue, Nov 27 2018 at 7:42pm -0500,
Post by Mikulas Patocka
Now when part_round_stats is gone, we can switch to per-cpu in-flight
counters.
We use the local-atomic type local_t, so that if part_inc_in_flight or
part_dec_in_flight is reentrantly called from an interrupt, the value will
be correct.
The other counters could be corrupted due to reentrant interrupt, but the
corruption only results in slight counter skew - the in_flight counter
must be exact, so it needs local_t.
---
block/bio.c | 4 ++--
block/blk-core.c | 4 ++--
block/blk-merge.c | 2 +-
block/genhd.c | 47 +++++++++++++++++++++++++++++++++++------------
drivers/md/dm.c | 4 +---
include/linux/genhd.h | 7 ++++---
6 files changed, 45 insertions(+), 23 deletions(-)
...
Post by Mikulas Patocka
Index: linux-block/drivers/md/dm.c
===================================================================
--- linux-block.orig/drivers/md/dm.c 2018-11-28 00:09:59.000000000 +0100
+++ linux-block/drivers/md/dm.c 2018-11-28 00:09:59.000000000 +0100
@@ -663,8 +663,7 @@ static void start_io_acct(struct dm_io *
generic_start_io_acct(md->queue, bio_op(bio), bio_sectors(bio),
&dm_disk(md)->part0);
- atomic_set(&dm_disk(md)->part0.in_flight[rw],
- atomic_inc_return(&md->pending[rw]));
+ atomic_inc(&md->pending[rw]);
if (unlikely(dm_stats_used(&md->stats)))
dm_stats_account_io(&md->stats, bio_data_dir(bio),
@@ -693,7 +692,6 @@ static void end_io_acct(struct dm_io *io
* a flush.
*/
pending = atomic_dec_return(&md->pending[rw]);
- atomic_set(&dm_disk(md)->part0.in_flight[rw], pending);
pending += atomic_read(&md->pending[rw^0x1]);
/* nudge anyone waiting on suspend queue */
These dm.c hunks conflict with changes from you that I already staged in
dm-4.21, see:

https://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm.git/commit/?h=dm-4.21&id=b5616f7a11592cc74860f4ec3e3c4fba6688eefa
https://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm.git/commit/?h=dm-4.21&id=78f95b53c203c969bbe6b86e405f7a891a43b6be

I'd really like to get away from DM maintaining its own ->pending
counters.

Mike
Mikulas Patocka
2018-11-29 22:05:37 UTC
Permalink
Post by Mike Snitzer
On Tue, Nov 27 2018 at 7:42pm -0500,
Post by Mikulas Patocka
Now when part_round_stats is gone, we can switch to per-cpu in-flight
counters.
We use the local-atomic type local_t, so that if part_inc_in_flight or
part_dec_in_flight is reentrantly called from an interrupt, the value will
be correct.
The other counters could be corrupted due to reentrant interrupt, but the
corruption only results in slight counter skew - the in_flight counter
must be exact, so it needs local_t.
---
block/bio.c | 4 ++--
block/blk-core.c | 4 ++--
block/blk-merge.c | 2 +-
block/genhd.c | 47 +++++++++++++++++++++++++++++++++++------------
drivers/md/dm.c | 4 +---
include/linux/genhd.h | 7 ++++---
6 files changed, 45 insertions(+), 23 deletions(-)
...
Post by Mikulas Patocka
Index: linux-block/drivers/md/dm.c
===================================================================
--- linux-block.orig/drivers/md/dm.c 2018-11-28 00:09:59.000000000 +0100
+++ linux-block/drivers/md/dm.c 2018-11-28 00:09:59.000000000 +0100
@@ -663,8 +663,7 @@ static void start_io_acct(struct dm_io *
generic_start_io_acct(md->queue, bio_op(bio), bio_sectors(bio),
&dm_disk(md)->part0);
- atomic_set(&dm_disk(md)->part0.in_flight[rw],
- atomic_inc_return(&md->pending[rw]));
+ atomic_inc(&md->pending[rw]);
if (unlikely(dm_stats_used(&md->stats)))
dm_stats_account_io(&md->stats, bio_data_dir(bio),
@@ -693,7 +692,6 @@ static void end_io_acct(struct dm_io *io
* a flush.
*/
pending = atomic_dec_return(&md->pending[rw]);
- atomic_set(&dm_disk(md)->part0.in_flight[rw], pending);
pending += atomic_read(&md->pending[rw^0x1]);
/* nudge anyone waiting on suspend queue */
These dm.c hunks conflict with changes from you that I already staged in
https://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm.git/commit/?h=dm-4.21&id=b5616f7a11592cc74860f4ec3e3c4fba6688eefa
https://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm.git/commit/?h=dm-4.21&id=78f95b53c203c969bbe6b86e405f7a891a43b6be
I'd really like to get away from DM maintaining its own ->pending
counters.
Mike
I know.

It depends on whether Jens takes these patches or not. If he doesn't take
them, dm will use percpu counters on its own and it will return zero in
the "/sys/block/dm-*/inflight" file.

Mikulas
Jens Axboe
2018-11-29 22:22:02 UTC
Permalink
Post by Mikulas Patocka
Post by Mike Snitzer
On Tue, Nov 27 2018 at 7:42pm -0500,
Post by Mikulas Patocka
Now when part_round_stats is gone, we can switch to per-cpu in-flight
counters.
We use the local-atomic type local_t, so that if part_inc_in_flight or
part_dec_in_flight is reentrantly called from an interrupt, the value will
be correct.
The other counters could be corrupted due to reentrant interrupt, but the
corruption only results in slight counter skew - the in_flight counter
must be exact, so it needs local_t.
---
block/bio.c | 4 ++--
block/blk-core.c | 4 ++--
block/blk-merge.c | 2 +-
block/genhd.c | 47 +++++++++++++++++++++++++++++++++++------------
drivers/md/dm.c | 4 +---
include/linux/genhd.h | 7 ++++---
6 files changed, 45 insertions(+), 23 deletions(-)
...
Post by Mikulas Patocka
Index: linux-block/drivers/md/dm.c
===================================================================
--- linux-block.orig/drivers/md/dm.c 2018-11-28 00:09:59.000000000 +0100
+++ linux-block/drivers/md/dm.c 2018-11-28 00:09:59.000000000 +0100
@@ -663,8 +663,7 @@ static void start_io_acct(struct dm_io *
generic_start_io_acct(md->queue, bio_op(bio), bio_sectors(bio),
&dm_disk(md)->part0);
- atomic_set(&dm_disk(md)->part0.in_flight[rw],
- atomic_inc_return(&md->pending[rw]));
+ atomic_inc(&md->pending[rw]);
if (unlikely(dm_stats_used(&md->stats)))
dm_stats_account_io(&md->stats, bio_data_dir(bio),
@@ -693,7 +692,6 @@ static void end_io_acct(struct dm_io *io
* a flush.
*/
pending = atomic_dec_return(&md->pending[rw]);
- atomic_set(&dm_disk(md)->part0.in_flight[rw], pending);
pending += atomic_read(&md->pending[rw^0x1]);
/* nudge anyone waiting on suspend queue */
These dm.c hunks conflict with changes from you that I already staged in
https://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm.git/commit/?h=dm-4.21&id=b5616f7a11592cc74860f4ec3e3c4fba6688eefa
https://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm.git/commit/?h=dm-4.21&id=78f95b53c203c969bbe6b86e405f7a891a43b6be
I'd really like to get away from DM maintaining its own ->pending
counters.
Mike
I know.
It depends on whether Jens takes these patches or not. If he doesn't take
them, dm will use percpu counters on its own and it will return zero in
the "/sys/block/dm-*/inflight" file.
This reply is the first I've seen of them?!

Oh, they are in gmail spam. This happened with your last patches as well,
you might want to look into that. I'll take a look at them later today.
--
Jens Axboe
Mike Snitzer
2018-11-30 00:21:48 UTC
Permalink
On Thu, Nov 29 2018 at 5:05pm -0500,
Post by Mikulas Patocka
Post by Mike Snitzer
On Tue, Nov 27 2018 at 7:42pm -0500,
Post by Mikulas Patocka
Now when part_round_stats is gone, we can switch to per-cpu in-flight
counters.
We use the local-atomic type local_t, so that if part_inc_in_flight or
part_dec_in_flight is reentrantly called from an interrupt, the value will
be correct.
The other counters could be corrupted due to reentrant interrupt, but the
corruption only results in slight counter skew - the in_flight counter
must be exact, so it needs local_t.
---
block/bio.c | 4 ++--
block/blk-core.c | 4 ++--
block/blk-merge.c | 2 +-
block/genhd.c | 47 +++++++++++++++++++++++++++++++++++------------
drivers/md/dm.c | 4 +---
include/linux/genhd.h | 7 ++++---
6 files changed, 45 insertions(+), 23 deletions(-)
...
Post by Mikulas Patocka
Index: linux-block/drivers/md/dm.c
===================================================================
--- linux-block.orig/drivers/md/dm.c 2018-11-28 00:09:59.000000000 +0100
+++ linux-block/drivers/md/dm.c 2018-11-28 00:09:59.000000000 +0100
@@ -663,8 +663,7 @@ static void start_io_acct(struct dm_io *
generic_start_io_acct(md->queue, bio_op(bio), bio_sectors(bio),
&dm_disk(md)->part0);
- atomic_set(&dm_disk(md)->part0.in_flight[rw],
- atomic_inc_return(&md->pending[rw]));
+ atomic_inc(&md->pending[rw]);
if (unlikely(dm_stats_used(&md->stats)))
dm_stats_account_io(&md->stats, bio_data_dir(bio),
@@ -693,7 +692,6 @@ static void end_io_acct(struct dm_io *io
* a flush.
*/
pending = atomic_dec_return(&md->pending[rw]);
- atomic_set(&dm_disk(md)->part0.in_flight[rw], pending);
pending += atomic_read(&md->pending[rw^0x1]);
/* nudge anyone waiting on suspend queue */
These dm.c hunks conflict with changes from you that I already staged in
https://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm.git/commit/?h=dm-4.21&id=b5616f7a11592cc74860f4ec3e3c4fba6688eefa
https://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm.git/commit/?h=dm-4.21&id=78f95b53c203c969bbe6b86e405f7a891a43b6be
I'd really like to get away from DM maintaining its own ->pending
counters.
Mike
I know.
It depends on whether Jens takes these patches or not. If he doesn't take
them, dm will use percpu counters on its own and it will return zero in
the "/sys/block/dm-*/inflight" file.
I see, yeah, I could easily rebase the changes I referenced.

I had a quick look at your proposed changes, they seem pretty clean to
me. The fact that disk_stats is already percpu for other counters makes
this change straight forward.

But I'll review closer.

Mike
Mike Snitzer
2018-11-30 21:44:33 UTC
Permalink
On Tue, Nov 27 2018 at 7:42pm -0500,
Post by Mikulas Patocka
Now when part_round_stats is gone, we can switch to per-cpu in-flight
counters.
We use the local-atomic type local_t, so that if part_inc_in_flight or
part_dec_in_flight is reentrantly called from an interrupt, the value will
be correct.
The other counters could be corrupted due to reentrant interrupt, but the
corruption only results in slight counter skew - the in_flight counter
must be exact, so it needs local_t.
---
block/bio.c | 4 ++--
block/blk-core.c | 4 ++--
block/blk-merge.c | 2 +-
block/genhd.c | 47 +++++++++++++++++++++++++++++++++++------------
drivers/md/dm.c | 4 +---
include/linux/genhd.h | 7 ++++---
6 files changed, 45 insertions(+), 23 deletions(-)
Index: linux-block/block/genhd.c
===================================================================
--- linux-block.orig/block/genhd.c 2018-11-28 00:09:59.000000000 +0100
+++ linux-block/block/genhd.c 2018-11-28 00:09:59.000000000 +0100
@@ -45,53 +45,76 @@ static void disk_add_events(struct gendi
static void disk_del_events(struct gendisk *disk);
static void disk_release_events(struct gendisk *disk);
-void part_inc_in_flight(struct request_queue *q, struct hd_struct *part, int rw)
+void part_inc_in_flight(struct request_queue *q, int cpu, struct hd_struct *part, int rw)
{
if (queue_is_mq(q))
return;
- atomic_inc(&part->in_flight[rw]);
+ local_inc(per_cpu_ptr(part->dkstats, cpu)->in_flight);
if (part->partno)
- atomic_inc(&part_to_disk(part)->part0.in_flight[rw]);
+ local_inc(per_cpu_ptr(part_to_disk(part)->part0.dkstats, cpu)->in_flight);
}
-void part_dec_in_flight(struct request_queue *q, struct hd_struct *part, int rw)
+void part_dec_in_flight(struct request_queue *q, int cpu, struct hd_struct *part, int rw)
{
if (queue_is_mq(q))
return;
- atomic_dec(&part->in_flight[rw]);
+ local_dec(per_cpu_ptr(part->dkstats, cpu)->in_flight);
if (part->partno)
- atomic_dec(&part_to_disk(part)->part0.in_flight[rw]);
+ local_dec(per_cpu_ptr(part_to_disk(part)->part0.dkstats, cpu)->in_flight);
}
The above is wrong, needs this incremental fix:

diff --git a/block/genhd.c b/block/genhd.c
index 2ed431098953..3397288a2926 100644
--- a/block/genhd.c
+++ b/block/genhd.c
@@ -50,9 +50,9 @@ void part_inc_in_flight(struct request_queue *q, int cpu, struct hd_struct *part
if (queue_is_mq(q))
return;

- local_inc(per_cpu_ptr(part->dkstats, cpu)->in_flight);
+ local_inc(&per_cpu_ptr(part->dkstats, cpu)->in_flight[rw]);
if (part->partno)
- local_inc(per_cpu_ptr(part_to_disk(part)->part0.dkstats, cpu)->in_flight);
+ local_inc(&per_cpu_ptr(part_to_disk(part)->part0.dkstats, cpu)->in_flight[rw]);
}

void part_dec_in_flight(struct request_queue *q, int cpu, struct hd_struct *part, int rw)
@@ -60,9 +60,9 @@ void part_dec_in_flight(struct request_queue *q, int cpu, struct hd_struct *part
if (queue_is_mq(q))
return;

- local_dec(per_cpu_ptr(part->dkstats, cpu)->in_flight);
+ local_dec(&per_cpu_ptr(part->dkstats, cpu)->in_flight[rw]);
if (part->partno)
- local_dec(per_cpu_ptr(part_to_disk(part)->part0.dkstats, cpu)->in_flight);
+ local_dec(&per_cpu_ptr(part_to_disk(part)->part0.dkstats, cpu)->in_flight[rw]);
}

unsigned int part_in_flight(struct request_queue *q, struct hd_struct *part)
Jens Axboe
2018-11-30 21:52:16 UTC
Permalink
Post by Mike Snitzer
On Tue, Nov 27 2018 at 7:42pm -0500,
Post by Mikulas Patocka
Now when part_round_stats is gone, we can switch to per-cpu in-flight
counters.
We use the local-atomic type local_t, so that if part_inc_in_flight or
part_dec_in_flight is reentrantly called from an interrupt, the value will
be correct.
The other counters could be corrupted due to reentrant interrupt, but the
corruption only results in slight counter skew - the in_flight counter
must be exact, so it needs local_t.
---
block/bio.c | 4 ++--
block/blk-core.c | 4 ++--
block/blk-merge.c | 2 +-
block/genhd.c | 47 +++++++++++++++++++++++++++++++++++------------
drivers/md/dm.c | 4 +---
include/linux/genhd.h | 7 ++++---
6 files changed, 45 insertions(+), 23 deletions(-)
Index: linux-block/block/genhd.c
===================================================================
--- linux-block.orig/block/genhd.c 2018-11-28 00:09:59.000000000 +0100
+++ linux-block/block/genhd.c 2018-11-28 00:09:59.000000000 +0100
@@ -45,53 +45,76 @@ static void disk_add_events(struct gendi
static void disk_del_events(struct gendisk *disk);
static void disk_release_events(struct gendisk *disk);
-void part_inc_in_flight(struct request_queue *q, struct hd_struct *part, int rw)
+void part_inc_in_flight(struct request_queue *q, int cpu, struct hd_struct *part, int rw)
{
if (queue_is_mq(q))
return;
- atomic_inc(&part->in_flight[rw]);
+ local_inc(per_cpu_ptr(part->dkstats, cpu)->in_flight);
if (part->partno)
- atomic_inc(&part_to_disk(part)->part0.in_flight[rw]);
+ local_inc(per_cpu_ptr(part_to_disk(part)->part0.dkstats, cpu)->in_flight);
}
-void part_dec_in_flight(struct request_queue *q, struct hd_struct *part, int rw)
+void part_dec_in_flight(struct request_queue *q, int cpu, struct hd_struct *part, int rw)
{
if (queue_is_mq(q))
return;
- atomic_dec(&part->in_flight[rw]);
+ local_dec(per_cpu_ptr(part->dkstats, cpu)->in_flight);
if (part->partno)
- atomic_dec(&part_to_disk(part)->part0.in_flight[rw]);
+ local_dec(per_cpu_ptr(part_to_disk(part)->part0.dkstats, cpu)->in_flight);
}
Should also be using this_cpu_inc/dec instead.
--
Jens Axboe
Mike Snitzer
2018-11-30 22:04:23 UTC
Permalink
On Fri, Nov 30 2018 at 4:52pm -0500,
Post by Jens Axboe
Post by Mike Snitzer
On Tue, Nov 27 2018 at 7:42pm -0500,
Post by Mikulas Patocka
Now when part_round_stats is gone, we can switch to per-cpu in-flight
counters.
We use the local-atomic type local_t, so that if part_inc_in_flight or
part_dec_in_flight is reentrantly called from an interrupt, the value will
be correct.
The other counters could be corrupted due to reentrant interrupt, but the
corruption only results in slight counter skew - the in_flight counter
must be exact, so it needs local_t.
---
block/bio.c | 4 ++--
block/blk-core.c | 4 ++--
block/blk-merge.c | 2 +-
block/genhd.c | 47 +++++++++++++++++++++++++++++++++++------------
drivers/md/dm.c | 4 +---
include/linux/genhd.h | 7 ++++---
6 files changed, 45 insertions(+), 23 deletions(-)
Index: linux-block/block/genhd.c
===================================================================
--- linux-block.orig/block/genhd.c 2018-11-28 00:09:59.000000000 +0100
+++ linux-block/block/genhd.c 2018-11-28 00:09:59.000000000 +0100
@@ -45,53 +45,76 @@ static void disk_add_events(struct gendi
static void disk_del_events(struct gendisk *disk);
static void disk_release_events(struct gendisk *disk);
-void part_inc_in_flight(struct request_queue *q, struct hd_struct *part, int rw)
+void part_inc_in_flight(struct request_queue *q, int cpu, struct hd_struct *part, int rw)
{
if (queue_is_mq(q))
return;
- atomic_inc(&part->in_flight[rw]);
+ local_inc(per_cpu_ptr(part->dkstats, cpu)->in_flight);
if (part->partno)
- atomic_inc(&part_to_disk(part)->part0.in_flight[rw]);
+ local_inc(per_cpu_ptr(part_to_disk(part)->part0.dkstats, cpu)->in_flight);
}
-void part_dec_in_flight(struct request_queue *q, struct hd_struct *part, int rw)
+void part_dec_in_flight(struct request_queue *q, int cpu, struct hd_struct *part, int rw)
{
if (queue_is_mq(q))
return;
- atomic_dec(&part->in_flight[rw]);
+ local_dec(per_cpu_ptr(part->dkstats, cpu)->in_flight);
if (part->partno)
- atomic_dec(&part_to_disk(part)->part0.in_flight[rw]);
+ local_dec(per_cpu_ptr(part_to_disk(part)->part0.dkstats, cpu)->in_flight);
}
Should also be using this_cpu_inc/dec instead.
How so? Mikulas is using local_t (original header speaks to why)
Loading...