blk-throttle: Avoid tracking latency if low limit is invalid
authorBaolin Wang <baolin.wang@linux.alibaba.com>
Thu, 8 Oct 2020 03:52:24 +0000 (11:52 +0800)
committerJens Axboe <axboe@kernel.dk>
Thu, 8 Oct 2020 14:01:37 +0000 (08:01 -0600)
The IO latency tracking is only for LOW limit, so we should add a
validation to avoid redundant latency tracking if the LOW limit
is not valid.

Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
block/blk-throttle.c

index 7e72102..b0d8f7c 100644 (file)
@@ -2100,7 +2100,7 @@ static void throtl_update_latency_buckets(struct throtl_data *td)
        unsigned long last_latency[2] = { 0 };
        unsigned long latency[2];
 
-       if (!blk_queue_nonrot(td->queue))
+       if (!blk_queue_nonrot(td->queue) || !td->limit_valid[LIMIT_LOW])
                return;
        if (time_before(jiffies, td->last_calculate_time + HZ))
                return;
@@ -2338,6 +2338,8 @@ void blk_throtl_bio_endio(struct bio *bio)
        if (!blkg)
                return;
        tg = blkg_to_tg(blkg);
+       if (!tg->td->limit_valid[LIMIT_LOW])
+               return;
 
        finish_time_ns = ktime_get_ns();
        tg->last_finish_time = finish_time_ns >> 10;