On 4/18/21 11:09 PM, Junxiao Bi wrote:
- finish_wait(&rqw->wait, &data.wq);
+ mutex_lock(&rqw->throttle_mutex);
+ wait_event(rqw->wait, acquire_inflight_cb(rqw, private_data));
+ mutex_unlock(&rqw->throttle_mutex);
This will break the throttle? There is a inflight io limitation. With
On 4/18/21 5:33 AM, Hillf Danton wrote:
On Sat, 17 Apr 2021 14:37:57 Junxiao Bi wrote:
On 4/17/21 3:10 AM, Hillf Danton wrote:
+ if (acquire_inflight_cb(rqw, private_data))
This function is to increase atomic variable rq_wait->inflight.
You are right.
What's the mutex for?
It cut
On 4/17/21 3:10 AM, Hillf Danton wrote:
--- a/block/blk-rq-qos.c
+++ b/block/blk-rq-qos.c
@@ -260,19 +260,17 @@ void rq_qos_wait(struct rq_wait *rqw, void *private_data,
.cb = acquire_inflight_cb,
.private_data = private_data,
};
- bool has_sleeper;
On 4/14/21 9:11 PM, Hillf Danton wrote:
On Wed, 14 Apr 2021 14:18:30 Junxiao Bi wrote:
There is a race bug which can cause io hung when multiple processes
run parallel in rq_qos_wait().
Let assume there were 4 processes P1/P2/P3/P4, P1/P2 were at the entry
of rq_qos_wait, and P3/P4 were waiting
There is a race bug which can cause io hung when multiple processes
run parallel in rq_qos_wait().
Let assume there were 4 processes P1/P2/P3/P4, P1/P2 were at the entry
of rq_qos_wait, and P3/P4 were waiting for io done, 2 io were inflight,
the inflight io limit was 2. See race below.
void rq_qos
5 matches
Mail list logo
