[PATCH] mac80211: minstrel: Scale expected_throughput by 10

[PATCH] mac80211: minstrel: Scale expected_throughput by 10

[Sven Eckelmann]

The change from cur_tp to the function
minstrel_get_tp_avg/minstrel_ht_get_tp_avg changed the unit used for the
current throughtput. For example in minstrel_ht the correct
conversion between them would be:

    mrs->cur_tp / 10 == minstrel_ht_get_tp_avg(..).

This factor 10 must also be included in the calculation of
minstrel_get_expected_throughput and minstrel_ht_get_expected_throughput to
get similar results as before the change.

Fixes: 6a27b2c40b48 ("mac80211: restructure per-rate throughput calculation into function")
Signed-off-by: Sven Eckelmann <sven@open-mesh.com>
---
 net/mac80211/rc80211_minstrel.c    | 2 +-
 net/mac80211/rc80211_minstrel_ht.c | 3 ++-
 2 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/net/mac80211/rc80211_minstrel.c b/net/mac80211/rc80211_minstrel.c
index 3ece7d1..b54f398 100644
--- a/net/mac80211/rc80211_minstrel.c
+++ b/net/mac80211/rc80211_minstrel.c
@@ -711,7 +711,7 @@ static u32 minstrel_get_expected_throughput(void *priv_sta)
 	 * computing cur_tp
 	 */
 	tmp_mrs = &mi->r[idx].stats;
-	tmp_cur_tp = minstrel_get_tp_avg(&mi->r[idx], tmp_mrs->prob_ewma);
+	tmp_cur_tp = minstrel_get_tp_avg(&mi->r[idx], tmp_mrs->prob_ewma) * 10;
 	tmp_cur_tp = tmp_cur_tp * 1200 * 8 / 1024;
 
 	return tmp_cur_tp;
diff --git a/net/mac80211/rc80211_minstrel_ht.c b/net/mac80211/rc80211_minstrel_ht.c
index 3928dbd..8928151 100644
--- a/net/mac80211/rc80211_minstrel_ht.c
+++ b/net/mac80211/rc80211_minstrel_ht.c
@@ -1334,7 +1334,8 @@ static u32 minstrel_ht_get_expected_throughput(void *priv_sta)
 	prob = mi->groups[i].rates[j].prob_ewma;
 
 	/* convert tp_avg from pkt per second in kbps */
-	tp_avg = minstrel_ht_get_tp_avg(mi, i, j, prob) * AVG_PKT_SIZE * 8 / 1024;
+	tp_avg = minstrel_ht_get_tp_avg(mi, i, j, prob) * 10;
+	tp_avg *= AVG_PKT_SIZE * 8 / 1024;
 
 	return tp_avg;
 }

Re: [PATCH] mac80211: minstrel: Scale expected_throughput by 10

[Johannes Berg]

On Mon, 2016-02-01 at 13:24 +0100, Sven Eckelmann wrote:
> The change from cur_tp to the function
> minstrel_get_tp_avg/minstrel_ht_get_tp_avg changed the unit used for
> the
> current throughtput. For example in minstrel_ht the correct
> conversion between them would be:
> 
>     mrs->cur_tp / 10 == minstrel_ht_get_tp_avg(..).
> 
> This factor 10 must also be included in the calculation of
> minstrel_get_expected_throughput and
> minstrel_ht_get_expected_throughput to
> get similar results as before the change.
> 

10 is a pretty expensive factor, perhaps that should use 16 instead?

johannes

Re: [PATCH] mac80211: minstrel: Scale expected_throughput by 10

[Sven Eckelmann]

[-- Attachment #1: Type: text/plain, Size: 747 bytes --]

On Monday 01 February 2016 13:30:37 Johannes Berg wrote:
> On Mon, 2016-02-01 at 13:24 +0100, Sven Eckelmann wrote:
> > The change from cur_tp to the function
> > minstrel_get_tp_avg/minstrel_ht_get_tp_avg changed the unit used for
> > the
> > current throughtput. For example in minstrel_ht the correct
> > conversion between them would be:
> > 
> >     mrs->cur_tp / 10 == minstrel_ht_get_tp_avg(..).
> > 
> > This factor 10 must also be included in the calculation of
> > minstrel_get_expected_throughput and
> > minstrel_ht_get_expected_throughput to
> > get similar results as before the change.
> > 
> 
> 10 is a pretty expensive factor, perhaps that should use 16 instead?

Not really funny but I will change the title.

Kind regards,
	Sven

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

Re: [PATCH] mac80211: minstrel: Scale expected_throughput by 10

[Johannes Berg]

On Mon, 2016-02-01 at 13:56 +0100, Sven Eckelmann wrote:
> On Monday 01 February 2016 13:30:37 Johannes Berg wrote:
> > On Mon, 2016-02-01 at 13:24 +0100, Sven Eckelmann wrote:
> > > The change from cur_tp to the function
> > > minstrel_get_tp_avg/minstrel_ht_get_tp_avg changed the unit used
> > > for
> > > the
> > > current throughtput. For example in minstrel_ht the correct
> > > conversion between them would be:
> > > 
> > >     mrs->cur_tp / 10 == minstrel_ht_get_tp_avg(..).
> > > 
> > > This factor 10 must also be included in the calculation of
> > > minstrel_get_expected_throughput and
> > > minstrel_ht_get_expected_throughput to
> > > get similar results as before the change.
> > > 
> > 
> > 10 is a pretty expensive factor, perhaps that should use 16
> > instead?
> 
> Not really funny but I will change the title.

Huh? Not sure what you mean. I really meant to change 10 to 16 overall
as far as the factor is concerned, since division by/multiplication
with 16 is far easier in base 2 than by/with 10.

johannes

Re: [PATCH] mac80211: minstrel: Scale expected_throughput by 10

[Antonio Quartulli]

[-- Attachment #1: Type: text/plain, Size: 1533 bytes --]

On Mon, Feb 01, 2016 at 01:57:53PM +0100, Johannes Berg wrote:
> On Mon, 2016-02-01 at 13:56 +0100, Sven Eckelmann wrote:
> > On Monday 01 February 2016 13:30:37 Johannes Berg wrote:
> > > On Mon, 2016-02-01 at 13:24 +0100, Sven Eckelmann wrote:
> > > > The change from cur_tp to the function
> > > > minstrel_get_tp_avg/minstrel_ht_get_tp_avg changed the unit used
> > > > for
> > > > the
> > > > current throughtput. For example in minstrel_ht the correct
> > > > conversion between them would be:
> > > > 
> > > >     mrs->cur_tp / 10 == minstrel_ht_get_tp_avg(..).
> > > > 
> > > > This factor 10 must also be included in the calculation of
> > > > minstrel_get_expected_throughput and
> > > > minstrel_ht_get_expected_throughput to
> > > > get similar results as before the change.
> > > > 
> > > 
> > > 10 is a pretty expensive factor, perhaps that should use 16
> > > instead?
> > 
> > Not really funny but I will change the title.
> 
> Huh? Not sure what you mean. I really meant to change 10 to 16 overall
> as far as the factor is concerned, since division by/multiplication
> with 16 is far easier in base 2 than by/with 10.

Should we first fix the bug introduced by 6a27b2c40b48 and then (with another
patch) improve this by changing the factor from 10 to 16 ?
The latter is going to be a bigger change because we need to make sure that the
value exported via station_info gets still scaled to kbps.

Cheers,


-- 
Antonio Quartulli
Director of Embedded Software | Open Mesh, Inc.

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

Re: [PATCH] mac80211: minstrel: Scale expected_throughput by 10

[Johannes Berg]

On Mon, 2016-02-01 at 20:58 +0800, Antonio Quartulli wrote:
> 
> Should we first fix the bug introduced by 6a27b2c40b48 and then (with
> another
> patch) improve this by changing the factor from 10 to 16 ?
> The latter is going to be a bigger change because we need to make
> sure that the
> value exported via station_info gets still scaled to kbps.
> 

Ok, I guess that makes sense.

johannes