cgroup: blkio: adjust block io bps limit by weight

Kernel now supports cgroup bps and iops hard limit by io-throttling.
But that's not a scalable way, because block devices' bandwith is
always changing over time.

This patch dynamically adjusts bps limit of each blkio cgroup according
to its pre-set weight. Additionally, if a process has not issued any io
request during a time window, its weight will be shared by other processes
until its next io request.

Signed-off-by: Lei Chen <lennychen@tencent.com>

block/blk-core.c

@@ -1514,6 +1514,19 @@ static void part_round_stats_single(struct request_queue *q, int cpu,
 	part->stamp = now;
 }
 
+static void part_round_stats_single_ns(struct request_queue *q, int cpu,
+				struct hd_struct *part, u64 now, 
+				unsigned int inflight)
+{
+	if (now == part->stamp_ns)
+		return;
+
+	if (inflight)
+		__part_stat_add(cpu, part, io_ticks_ns, (now - part->stamp_ns));
+	part->stamp_ns = now;
+}
+
+
 /**
  * part_round_stats() - Round off the performance stats on a struct disk_stats.
  * @q: target block queue
@@ -1537,6 +1550,7 @@ void part_round_stats(struct request_queue *q, int cpu, struct hd_struct *part)
 	unsigned long now = jiffies;
 	unsigned int inflight[2];
 	int stats = 0;
+	u64 now_ns = ktime_to_ns(ktime_get());
 
 	if (part->stamp != now)
 		stats |= 1;
@@ -1552,10 +1566,17 @@ void part_round_stats(struct request_queue *q, int cpu, struct hd_struct *part)
 
 	part_in_flight(q, part, inflight);
 
-	if (stats & 2)
+	if (stats & 2) {
+		/*for sda1 sda2 sda3*/
 		part_round_stats_single(q, cpu, part2, now, inflight[1]);
-	if (stats & 1)
+		part_round_stats_single_ns(q, cpu, part2, now_ns, inflight[1]);
+	}
+
+	if (stats & 1) {
+		/*for sda*/
 		part_round_stats_single(q, cpu, part, now, inflight[0]);
+		part_round_stats_single_ns(q, cpu, part, now_ns, inflight[0]);
+	}
 }
 EXPORT_SYMBOL_GPL(part_round_stats);
 

block/blk-sysfs.c

@@ -364,6 +364,17 @@ queue_rq_affinity_store(struct request_queue *q, const char *page, size_t count)
 	return ret;
 }
 
+static ssize_t queue_avg_perf_show(struct request_queue *q, char *page)
+{
+	return sprintf(page, "%llu\n",
+		       (unsigned long long)q->disk_bw * 512);
+}
+
+static struct queue_sysfs_entry queue_avg_perf_entry = {
+	.attr = {.name = "average_perf", .mode = S_IRUGO },
+	.show = queue_avg_perf_show,
+};
+
 static ssize_t queue_poll_delay_show(struct request_queue *q, char *page)
 {
 	int val;
@@ -688,6 +699,7 @@ static struct queue_sysfs_entry throtl_sample_time_entry = {
 #endif
 
 static struct attribute *default_attrs[] = {
+	&queue_avg_perf_entry.attr,
 	&queue_requests_entry.attr,
 	&queue_ra_entry.attr,
 	&queue_max_hw_sectors_entry.attr,