lkml.org 
[lkml]   [2009]   [Jan]   [6]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
SubjectRe: [PATCH] Allow RT requests to pre-empt ongoing BE timeslice in CFQ.
On Mon, Jan 05 2009, Divyesh Shah wrote:
> This patch adds the ability to pre-empt an ongoing BE timeslice when a RT
> request is waiting for the current timeslice to complete. This reduces the
> wait time to disk for RT requests from an upper bound of 4 (current value
> of cfq_quantum) to 1 disk request.
>
> Signed-off-by: Divyesh Shah <dpshah@google.com>
> ---
> Patch based on 2.6.28 linus tree.
>
> Test Results:
> Latency(secs) for the RT task when doing sequential reads from 10G file.
> only RT | RT + BE | RT + BE + this patch
> small (512 byte) reads | 143 | 163 | 146
> large (1Mb) reads | 142 | 158 | 146

Results look good. I'm not too keen on doing the rb lookup, how about we
just track pending rt queues instead? Add a counter to cfqd
(busy_rq_queues) and inc/dec that along with busy_queues, then check
that in the two locations? If that value is non-zero, an rt queue should
be returned with cfq_rb_first() as well.

Also, the places that test cfq_class_be() seem more appropriate as
!cfq_class_rt() checks, as that also covers idle queues.

How does the below look to you? It's untested here, but it compiles. I'd
appreciate your comments and testing!

diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c
index e8525fa..8640199 100644
--- a/block/cfq-iosched.c
+++ b/block/cfq-iosched.c
@@ -84,6 +84,7 @@ struct cfq_data {
*/
struct cfq_rb_root service_tree;
unsigned int busy_queues;
+ unsigned int busy_rt_queues;

int rq_in_driver;
int sync_flight;
@@ -562,6 +563,8 @@ static void cfq_add_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
BUG_ON(cfq_cfqq_on_rr(cfqq));
cfq_mark_cfqq_on_rr(cfqq);
cfqd->busy_queues++;
+ if (cfq_class_rt(cfqq))
+ cfqd->busy_rt_queues++;

cfq_resort_rr_list(cfqd, cfqq);
}
@@ -581,6 +584,8 @@ static void cfq_del_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)

BUG_ON(!cfqd->busy_queues);
cfqd->busy_queues--;
+ if (cfq_class_rt(cfqq))
+ cfqd->busy_rt_queues--;
}

/*
@@ -1005,6 +1010,20 @@ static struct cfq_queue *cfq_select_queue(struct cfq_data *cfqd)
goto expire;

/*
+ * If we have RT queues waiting, then we preempt the current non-rt
+ * cfqq.
+ */
+ if (!cfq_class_rt(cfqq) && cfqd->busy_rt_queues) {
+ /*
+ * We simulate this as cfqq timed out so that it gets to bank
+ * the remaining of its time slice.
+ */
+ cfq_log_cfqq(cfqd, cfqq, "preempt from RT");
+ cfq_slice_expired(cfqd, 1);
+ goto new_queue;
+ }
+
+ /*
* The active queue has requests and isn't expired, allow it to
* dispatch.
*/
@@ -1067,6 +1086,12 @@ __cfq_dispatch_requests(struct cfq_data *cfqd, struct cfq_queue *cfqq,
if (RB_EMPTY_ROOT(&cfqq->sort_list))
break;

+ /*
+ * If we have pending rt queues, stop dispatching
+ */
+ if (!cfq_class_rt(cfqq) && cfqd->busy_rt_queues)
+ break;
+
} while (dispatched < max_dispatch);

/*
@@ -1801,6 +1826,12 @@ cfq_should_preempt(struct cfq_data *cfqd, struct cfq_queue *new_cfqq,
if (rq_is_meta(rq) && !cfqq->meta_pending)
return 1;

+ /*
+ * Allow an RT request to pre-empt an ongoing BE cfqq timeslice.
+ */
+ if (cfq_class_rt(new_cfqq) && !cfq_class_rt(cfqq))
+ return 1;
+
if (!cfqd->active_cic || !cfq_cfqq_wait_request(cfqq))
return 0;

--
Jens Axboe



\
 
 \ /
  Last update: 2009-01-06 10:03    [W:0.101 / U:0.268 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site