On 08/17/16 22:55, Sreekanth Reddy wrote:
> Observing softlockups while running heavy IOs on 8 SSD drives
> connected behind our LSI SAS 3004 HBA.
Hello Sreekanth,
This means that more than 23s was spent before the scheduler was
invoked, probably due to a loop. Can you give the attached (untested)
patch a try to see whether it is the loop in __blk_mq_run_hw_queue()?
Thanks,
Bart.
From 4da94f2ec37ee5d1b4a5f1ce2886bdafd5cd394c Mon Sep 17 00:00:00 2001
From: Bart Van Assche <bart.vanassche@xxxxxxxxxxx>
Date: Thu, 18 Aug 2016 07:51:49 -0700
Subject: [PATCH] block: Measure __blk_mq_run_hw_queue() execution time
Note: the "max_elapsed" variable can be modified by multiple threads
concurrently.
---
block/blk-mq.c | 10 ++++++++++
1 file changed, 10 insertions(+)
diff --git a/block/blk-mq.c b/block/blk-mq.c
index e931a0e..6d0961c 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -792,6 +792,9 @@ static void __blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx)
LIST_HEAD(driver_list);
struct list_head *dptr;
int queued;
+ static long max_elapsed = -1;
+ unsigned long start = jiffies;
+ long elapsed;
WARN_ON(!cpumask_test_cpu(raw_smp_processor_id(), hctx->cpumask));
@@ -889,6 +892,13 @@ static void __blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx)
**/
blk_mq_run_hw_queue(hctx, true);
}
+
+ elapsed = jiffies - start;
+ if (elapsed > max_elapsed) {
+ max_elapsed = elapsed;
+ pr_info("%s() finished after %d ms\n", __func__,
+ jiffies_to_msecs(elapsed));
+ }
}
/*
--
2.9.2
