[PATCH] cfq: Remove special treatment for metadata rqs.

[PATCH] cfq: Remove special treatment for metadata rqs.

[Justin TerAvest]

There is no consistency among filesystems from what bios (or requests)
are marked as being metadata. It's interesting to expose this in traces,
but we shouldn't schedule the requests differently based on whether or
not they're marked as being metadata.

Signed-off-by: Justin TerAvest <teravest@google.com>
---
 block/cfq-iosched.c |   15 ---------------
 1 files changed, 0 insertions(+), 15 deletions(-)

diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c
index 3d403a1..0079059 100644
--- a/block/cfq-iosched.c
+++ b/block/cfq-iosched.c
@@ -129,8 +129,6 @@ struct cfq_queue {
 	unsigned long slice_end;
 	long slice_resid;
 
-	/* pending metadata requests */
-	int meta_pending;
 	/* number of requests that are on the dispatch list or inside driver */
 	int dispatched;
 
@@ -1593,10 +1591,6 @@ static void cfq_remove_request(struct request *rq)
 	cfqq->cfqd->rq_queued--;
 	cfq_blkiocg_update_io_remove_stats(&(RQ_CFQG(rq))->blkg,
 					rq_data_dir(rq), rq_is_sync(rq));
-	if (rq->cmd_flags & REQ_META) {
-		WARN_ON(!cfqq->meta_pending);
-		cfqq->meta_pending--;
-	}
 }
 
 static int cfq_merge(struct request_queue *q, struct request **req,
@@ -3331,13 +3325,6 @@ cfq_should_preempt(struct cfq_data *cfqd, struct cfq_queue *new_cfqq,
 		return true;
 
 	/*
-	 * So both queues are sync. Let the new request get disk time if
-	 * it's a metadata request and the current queue is doing regular IO.
-	 */
-	if ((rq->cmd_flags & REQ_META) && !cfqq->meta_pending)
-		return true;
-
-	/*
 	 * Allow an RT request to pre-empt an ongoing non-RT cfqq timeslice.
 	 */
 	if (cfq_class_rt(new_cfqq) && !cfq_class_rt(cfqq))
@@ -3401,8 +3388,6 @@ cfq_rq_enqueued(struct cfq_data *cfqd, struct cfq_queue *cfqq,
 	struct cfq_io_context *cic = RQ_CIC(rq);
 
 	cfqd->rq_queued++;
-	if (rq->cmd_flags & REQ_META)
-		cfqq->meta_pending++;
 
 	cfq_update_io_thinktime(cfqd, cic);
 	cfq_update_io_seektime(cfqd, cfqq, rq);
-- 
1.7.3.1

Re: [PATCH] cfq: Remove special treatment for metadata rqs.

[Christoph Hellwig]

On Mon, Jun 20, 2011 at 12:52:39PM -0700, Justin TerAvest wrote:
> There is no consistency among filesystems from what bios (or requests)
> are marked as being metadata. It's interesting to expose this in traces,
> but we shouldn't schedule the requests differently based on whether or
> not they're marked as being metadata.

Personally I totally agree.  If anyone can demonstrate an effect from
the current hacks we shall add them back with a special REQ_ flag just
for the boost.

Re: [PATCH] cfq: Remove special treatment for metadata rqs.

[Jens Axboe]

On 2011-06-20 21:52, Justin TerAvest wrote:
> There is no consistency among filesystems from what bios (or requests)
> are marked as being metadata. It's interesting to expose this in traces,
> but we shouldn't schedule the requests differently based on whether or
> not they're marked as being metadata.

Agree, we should kill that. What is this patch against? cfq_choose_req()
also has a "choose meta over non-meta" preference.

-- 
Jens Axboe

Re: [PATCH] cfq: Remove special treatment for metadata rqs.

[Justin TerAvest]

On Mon, Jun 20, 2011 at 12:59 PM, Jens Axboe <jaxboe@fusionio.com> wrote:
> On 2011-06-20 21:52, Justin TerAvest wrote:
>> There is no consistency among filesystems from what bios (or requests)
>> are marked as being metadata. It's interesting to expose this in traces,
>> but we shouldn't schedule the requests differently based on whether or
>> not they're marked as being metadata.
>
> Agree, we should kill that. What is this patch against? cfq_choose_req()
> also has a "choose meta over non-meta" preference.

This is against for-3.1. Let me know if you want it against something else.

Sorry, I missed the preference in cfq_choose_req, since it changed to
be clever about bit operations. I'll send a v2 patch that cleans that
use up as well.

>
> --
> Jens Axboe
>
>

Re: [PATCH] cfq: Remove special treatment for metadata rqs.

[Jens Axboe]

On 2011-06-20 22:03, Justin TerAvest wrote:
> On Mon, Jun 20, 2011 at 12:59 PM, Jens Axboe <jaxboe@fusionio.com> wrote:
>> On 2011-06-20 21:52, Justin TerAvest wrote:
>>> There is no consistency among filesystems from what bios (or requests)
>>> are marked as being metadata. It's interesting to expose this in traces,
>>> but we shouldn't schedule the requests differently based on whether or
>>> not they're marked as being metadata.
>>
>> Agree, we should kill that. What is this patch against? cfq_choose_req()
>> also has a "choose meta over non-meta" preference.
> 
> This is against for-3.1. Let me know if you want it against something else.

Great, right branch.

> Sorry, I missed the preference in cfq_choose_req, since it changed to
> be clever about bit operations. I'll send a v2 patch that cleans that
> use up as well.

Thanks!

-- 
Jens Axboe