block: add iostat counters for flush requests

Requests that triggers flushing volatile writeback cache to disk (barriers)
have significant effect to overall performance.

Block layer has sophisticated engine for combining several flush requests
into one. But there is no statistics for actual flushes executed by disk.
Requests which trigger flushes usually are barriers - zero-size writes.

This patch adds two iostat counters into /sys/class/block/$dev/stat and
/proc/diskstats - count of completed flush requests and their total time.

Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Documentation/ABI/testing/procfs-diskstats

@@ -29,4 +29,9 @@ Description:
 		17 - sectors discarded
 		18 - time spent discarding
 
+		Kernel 5.5+ appends two more fields for flush requests:
+
+		19 - flush requests completed successfully
+		20 - time spent flushing
+
 		For more details refer to Documentation/admin-guide/iostats.rst

Documentation/ABI/testing/sysfs-block

@@ -15,6 +15,12 @@ Description:
 		 9 - I/Os currently in progress
 		10 - time spent doing I/Os (ms)
 		11 - weighted time spent doing I/Os (ms)
+		12 - discards completed
+		13 - discards merged
+		14 - sectors discarded
+		15 - time spent discarding (ms)
+		16 - flush requests completed
+		17 - time spent flushing (ms)
 		For more details refer Documentation/admin-guide/iostats.rst
 
 

Documentation/admin-guide/iostats.rst

@@ -121,6 +121,15 @@ Field 15 -- # of milliseconds spent discarding
     This is the total number of milliseconds spent by all discards (as
     measured from __make_request() to end_that_request_last()).
 
+Field 16 -- # of flush requests completed
+    This is the total number of flush requests completed successfully.
+
+    Block layer combines flush requests and executes at most one at a time.
+    This counts flush requests executed by disk. Not tracked for partitions.
+
+Field 17 -- # of milliseconds spent flushing
+    This is the total number of milliseconds spent by all flush requests.
+
 To avoid introducing performance bottlenecks, no locks are held while
 modifying these counters.  This implies that minor inaccuracies may be
 introduced when changes collide, so (for instance) adding up all the

Documentation/block/stat.rst

@@ -41,6 +41,8 @@ discard I/Os    requests      number of discard I/Os processed
 discard merges  requests      number of discard I/Os merged with in-queue I/O
 discard sectors sectors       number of sectors discarded
 discard ticks   milliseconds  total wait time for discard requests
+flush I/Os      requests      number of flush I/Os processed
+flush ticks     milliseconds  total wait time for flush requests
 =============== ============= =================================================
 
 read I/Os, write I/Os, discard I/0s
@@ -48,6 +50,14 @@ read I/Os, write I/Os, discard I/0s
 
 These values increment when an I/O request completes.
 
+flush I/Os
+==========
+
+These values increment when an flush I/O request completes.
+
+Block layer combines flush requests and executes at most one at a time.
+This counts flush requests executed by disk. Not tracked for partitions.
+
 read merges, write merges, discard merges
 =========================================
 
@@ -62,8 +72,8 @@ discarded from this block device.  The "sectors" in question are the
 standard UNIX 512-byte sectors, not any device- or filesystem-specific
 block size.  The counters are incremented when the I/O completes.
 
-read ticks, write ticks, discard ticks
-======================================
+read ticks, write ticks, discard ticks, flush ticks
+===================================================
 
 These values count the number of milliseconds that I/O requests have
 waited on this block device.  If there are multiple I/O requests waiting,

block/blk-flush.c

@@ -136,6 +136,17 @@ static void blk_flush_queue_rq(struct request *rq, bool add_front)
 	blk_mq_add_to_requeue_list(rq, add_front, true);
 }
 
+static void blk_account_io_flush(struct request *rq)
+{
+	struct hd_struct *part = &rq->rq_disk->part0;
+
+	part_stat_lock();
+	part_stat_inc(part, ios[STAT_FLUSH]);
+	part_stat_add(part, nsecs[STAT_FLUSH],
+		      ktime_get_ns() - rq->start_time_ns);
+	part_stat_unlock();
+}
+
 /**
  * blk_flush_complete_seq - complete flush sequence
  * @rq: PREFLUSH/FUA request being sequenced
@@ -185,7 +196,7 @@ static void blk_flush_complete_seq(struct request *rq,
 
 	case REQ_FSEQ_DONE:
 		/*
-		 * @rq was previously adjusted by blk_flush_issue() for
+		 * @rq was previously adjusted by blk_insert_flush() for
 		 * flush sequencing and may already have gone through the
 		 * flush data request completion path.  Restore @rq for
 		 * normal completion and end it.
@@ -212,6 +223,8 @@ static void flush_end_io(struct request *flush_rq, blk_status_t error)
 	struct blk_flush_queue *fq = blk_get_flush_queue(q, flush_rq->mq_ctx);
 	struct blk_mq_hw_ctx *hctx;
 
+	blk_account_io_flush(flush_rq);
+
 	/* release the tag's ownership to the req cloned from */
 	spin_lock_irqsave(&fq->mq_flush_lock, flags);
 

block/genhd.c

@@ -1385,7 +1385,9 @@ static int diskstats_show(struct seq_file *seqf, void *v)
 			   "%lu %lu %lu %u "
 			   "%lu %lu %lu %u "
 			   "%u %u %u "
-			   "%lu %lu %lu %u\n",
+			   "%lu %lu %lu %u "
+			   "%lu %u"
+			   "\n",
 			   MAJOR(part_devt(hd)), MINOR(part_devt(hd)),
 			   disk_name(gp, hd->partno, buf),
 			   part_stat_read(hd, ios[STAT_READ]),
@@ -1402,7 +1404,9 @@ static int diskstats_show(struct seq_file *seqf, void *v)
 			   part_stat_read(hd, ios[STAT_DISCARD]),
 			   part_stat_read(hd, merges[STAT_DISCARD]),
 			   part_stat_read(hd, sectors[STAT_DISCARD]),
-			   (unsigned int)part_stat_read_msecs(hd, STAT_DISCARD)
+			   (unsigned int)part_stat_read_msecs(hd, STAT_DISCARD),
+			   part_stat_read(hd, ios[STAT_FLUSH]),
+			   (unsigned int)part_stat_read_msecs(hd, STAT_FLUSH)
 			);
 	}
 	disk_part_iter_exit(&piter);

block/partition-generic.c

@@ -128,6 +128,7 @@ ssize_t part_stat_show(struct device *dev,
 		"%8lu %8lu %8llu %8u "
 		"%8u %8u %8u "
 		"%8lu %8lu %8llu %8u "
+		"%8lu %8u"
 		"\n",
 		part_stat_read(p, ios[STAT_READ]),
 		part_stat_read(p, merges[STAT_READ]),
@@ -143,7 +144,9 @@ ssize_t part_stat_show(struct device *dev,
 		part_stat_read(p, ios[STAT_DISCARD]),
 		part_stat_read(p, merges[STAT_DISCARD]),
 		(unsigned long long)part_stat_read(p, sectors[STAT_DISCARD]),
-		(unsigned int)part_stat_read_msecs(p, STAT_DISCARD));
+		(unsigned int)part_stat_read_msecs(p, STAT_DISCARD),
+		part_stat_read(p, ios[STAT_FLUSH]),
+		(unsigned int)part_stat_read_msecs(p, STAT_FLUSH));
 }
 
 ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,

include/linux/blk_types.h

@@ -377,6 +377,7 @@ enum stat_group {
 	STAT_READ,
 	STAT_WRITE,
 	STAT_DISCARD,
+	STAT_FLUSH,
 
 	NR_STAT_GROUPS
 };