block: Seperate read and write statistics of in_flight requests v2

Commit a9327cac440be4d8333bba975cbbf76045096275 added seperate read
and write statistics of in_flight requests. And exported the number
of read and write requests in progress seperately through sysfs.

But Corrado Zoccolo <[e-mail]> reported getting strange
output from "iostat -kx 2". Global values for service time and
utilization were garbage. For interval values, utilization was always
100%, and service time is higher than normal.

So this was reverted by commit 0f78ab9899e9d6acb09d5465def618704255963b

The problem was in part_round_stats_single(), I missed the following:
if (now == part->stamp)
return;

- if (part->in_flight) {
+ if (part_in_flight(part)) {
__part_stat_add(cpu, part, time_in_queue,
part_in_flight(part) * (now - part->stamp));
__part_stat_add(cpu, part, io_ticks, (now - part->stamp));

With this chunk included, the reported regression gets fixed.

Signed-off-by: Nikanth Karthikesan <[e-mail]>

--
Signed-off-by: Jens Axboe <[e-mail]>

block/blk-core.c

···
 		part_stat_inc(cpu, part, merges[rw]);
 	else {
 		part_round_stats(cpu, part);
-		part_inc_in_flight(part);
+		part_inc_in_flight(part, rw);
 	}
 
 	part_stat_unlock();
···
 	if (now == part->stamp)
 		return;
 
-	if (part->in_flight) {
+	if (part_in_flight(part)) {
 		__part_stat_add(cpu, part, time_in_queue,
-				part->in_flight * (now - part->stamp));
+				part_in_flight(part) * (now - part->stamp));
 		__part_stat_add(cpu, part, io_ticks, (now - part->stamp));
 	}
 	part->stamp = now;
···
 		part_stat_inc(cpu, part, ios[rw]);
 		part_stat_add(cpu, part, ticks[rw], duration);
 		part_round_stats(cpu, part);
-		part_dec_in_flight(part);
+		part_dec_in_flight(part, rw);
 
 		part_stat_unlock();
 	}

block/blk-merge.c

···
 		part = disk_map_sector_rcu(req->rq_disk, blk_rq_pos(req));
 
 		part_round_stats(cpu, part);
-		part_dec_in_flight(part);
+		part_dec_in_flight(part, rq_data_dir(req));
 
 		part_stat_unlock();
 	}

block/genhd.c

···
 static DEVICE_ATTR(alignment_offset, S_IRUGO, disk_alignment_offset_show, NULL);
 static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
 static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
+static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
 #ifdef CONFIG_FAIL_MAKE_REQUEST
 static struct device_attribute dev_attr_fail =
 	__ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
···
 	&dev_attr_alignment_offset.attr,
 	&dev_attr_capability.attr,
 	&dev_attr_stat.attr,
+	&dev_attr_inflight.attr,
 #ifdef CONFIG_FAIL_MAKE_REQUEST
 	&dev_attr_fail.attr,
 #endif
···
 			   part_stat_read(hd, merges[1]),
 			   (unsigned long long)part_stat_read(hd, sectors[1]),
 			   jiffies_to_msecs(part_stat_read(hd, ticks[1])),
-			   hd->in_flight,
+			   part_in_flight(hd),
 			   jiffies_to_msecs(part_stat_read(hd, io_ticks)),
 			   jiffies_to_msecs(part_stat_read(hd, time_in_queue))
 			);

drivers/md/dm.c

···
 	/*
 	 * A list of ios that arrived while we were suspended.
 	 */
-	atomic_t pending;
+	atomic_t pending[2];
 	wait_queue_head_t wait;
 	struct work_struct work;
 	struct bio_list deferred;
···
 {
 	struct mapped_device *md = io->md;
 	int cpu;
+	int rw = bio_data_dir(io->bio);
 
 	io->start_time = jiffies;
 
 	cpu = part_stat_lock();
 	part_round_stats(cpu, &dm_disk(md)->part0);
 	part_stat_unlock();
-	dm_disk(md)->part0.in_flight = atomic_inc_return(&md->pending);
+	dm_disk(md)->part0.in_flight[rw] = atomic_inc_return(&md->pending[rw]);
 }
 
 static void end_io_acct(struct dm_io *io)
···
 	 * After this is decremented the bio must not be touched if it is
 	 * a barrier.
 	 */
-	dm_disk(md)->part0.in_flight = pending =
-		atomic_dec_return(&md->pending);
+	dm_disk(md)->part0.in_flight[rw] = pending =
+		atomic_dec_return(&md->pending[rw]);
+	pending += atomic_read(&md->pending[rw^0x1]);
 
 	/* nudge anyone waiting on suspend queue */
 	if (!pending)
···
 	if (!md->disk)
 		goto bad_disk;
 
-	atomic_set(&md->pending, 0);
+	atomic_set(&md->pending[0], 0);
+	atomic_set(&md->pending[1], 0);
 	init_waitqueue_head(&md->wait);
 	INIT_WORK(&md->work, dm_wq_work);
 	init_waitqueue_head(&md->eventq);
···
 				break;
 			}
 			spin_unlock_irqrestore(q->queue_lock, flags);
-		} else if (!atomic_read(&md->pending))
+		} else if (!atomic_read(&md->pending[0]) &&
+					!atomic_read(&md->pending[1]))
 			break;
 
 		if (interruptible == TASK_INTERRUPTIBLE &&

fs/partitions/check.c

···
 		part_stat_read(p, merges[WRITE]),
 		(unsigned long long)part_stat_read(p, sectors[WRITE]),
 		jiffies_to_msecs(part_stat_read(p, ticks[WRITE])),
-		p->in_flight,
+		part_in_flight(p),
 		jiffies_to_msecs(part_stat_read(p, io_ticks)),
 		jiffies_to_msecs(part_stat_read(p, time_in_queue)));
+}
+
+ssize_t part_inflight_show(struct device *dev,
+			struct device_attribute *attr, char *buf)
+{
+	struct hd_struct *p = dev_to_part(dev);
+
+	return sprintf(buf, "%8u %8u\n", p->in_flight[0], p->in_flight[1]);
 }
 
 #ifdef CONFIG_FAIL_MAKE_REQUEST
···
 static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
 static DEVICE_ATTR(alignment_offset, S_IRUGO, part_alignment_offset_show, NULL);
 static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
+static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
 #ifdef CONFIG_FAIL_MAKE_REQUEST
 static struct device_attribute dev_attr_fail =
 	__ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
···
 	&dev_attr_size.attr,
 	&dev_attr_alignment_offset.attr,
 	&dev_attr_stat.attr,
+	&dev_attr_inflight.attr,
 #ifdef CONFIG_FAIL_MAKE_REQUEST
 	&dev_attr_fail.attr,
 #endif

include/linux/genhd.h

···
 	int make_it_fail;
 #endif
 	unsigned long stamp;
-	int in_flight;
+	int in_flight[2];
 #ifdef	CONFIG_SMP
 	struct disk_stats *dkstats;
 #else
···
 #define part_stat_sub(cpu, gendiskp, field, subnd)			\
 	part_stat_add(cpu, gendiskp, field, -subnd)
 
-static inline void part_inc_in_flight(struct hd_struct *part)
+static inline void part_inc_in_flight(struct hd_struct *part, int rw)
 {
-	part->in_flight++;
+	part->in_flight[rw]++;
 	if (part->partno)
-		part_to_disk(part)->part0.in_flight++;
+		part_to_disk(part)->part0.in_flight[rw]++;
 }
 
-static inline void part_dec_in_flight(struct hd_struct *part)
+static inline void part_dec_in_flight(struct hd_struct *part, int rw)
 {
-	part->in_flight--;
+	part->in_flight[rw]--;
 	if (part->partno)
-		part_to_disk(part)->part0.in_flight--;
+		part_to_disk(part)->part0.in_flight[rw]--;
+}
+
+static inline int part_in_flight(struct hd_struct *part)
+{
+	return part->in_flight[0] + part->in_flight[1];
 }
 
 /* block/blk-core.c */
···
 extern ssize_t part_size_show(struct device *dev,
 			      struct device_attribute *attr, char *buf);
 extern ssize_t part_stat_show(struct device *dev,
+			      struct device_attribute *attr, char *buf);
+extern ssize_t part_inflight_show(struct device *dev,
 			      struct device_attribute *attr, char *buf);
 #ifdef CONFIG_FAIL_MAKE_REQUEST
 extern ssize_t part_fail_show(struct device *dev,