[PATCH 1/2] mac80211: insert mesh peer after init

[PATCH 1/2] mac80211: insert mesh peer after init

[Thomas Pedersen]

Drivers need the station rate info when inserting a new sta_info. The
patch "mac80211: refactor mesh peer initialization" wrongly assumed the
rate info could be applied after insertion. After further review, this
is clearly not the case.

This fixes a regression where HT parameters were not applied before
inserting the sta_info, causing performance degradation.

Signed-off-by: Thomas Pedersen <thomas@cozybit.com>
---
 net/mac80211/mesh_plink.c |    8 +++++---
 1 files changed, 5 insertions(+), 3 deletions(-)

diff --git a/net/mac80211/mesh_plink.c b/net/mac80211/mesh_plink.c
index 1ff2a5c..f4124d7 100644
--- a/net/mac80211/mesh_plink.c
+++ b/net/mac80211/mesh_plink.c
@@ -102,9 +102,6 @@ static struct sta_info *mesh_plink_alloc(struct ieee80211_sub_if_data *sdata,
 
 	set_sta_flag(sta, WLAN_STA_WME);
 
-	if (sta_info_insert(sta))
-		return NULL;
-
 	return sta;
 }
 
@@ -281,6 +278,7 @@ static struct sta_info *mesh_peer_init(struct ieee80211_sub_if_data *sdata,
 	struct ieee80211_supported_band *sband;
 	u32 rates, basic_rates = 0;
 	struct sta_info *sta;
+	bool insert = false;
 
 	sband = local->hw.wiphy->bands[band];
 	rates = ieee80211_sta_get_rates(local, elems, band, &basic_rates);
@@ -290,6 +288,7 @@ static struct sta_info *mesh_peer_init(struct ieee80211_sub_if_data *sdata,
 		sta = mesh_plink_alloc(sdata, addr);
 		if (!sta)
 			return NULL;
+		insert = true;
 	}
 
 	spin_lock_bh(&sta->lock);
@@ -306,6 +305,9 @@ static struct sta_info *mesh_peer_init(struct ieee80211_sub_if_data *sdata,
 	rate_control_rate_init(sta);
 	spin_unlock_bh(&sta->lock);
 
+	if (insert && sta_info_insert(sta))
+		return NULL;
+
 	return sta;
 }
 
-- 
1.7.5.4

[PATCH 2/2] mac80211: don't transmit 40MHz frames to 20MHz peer

[Thomas Pedersen]

If a mesh peer indicates it is operating as 20MHz-only in its HT
operation IE, have the rate control algorithm respect this by disabling
the equivalent bit in the ieee80211_sta HT capabilities.

Signed-off-by: Thomas Pedersen <thomas@cozybit.com>
---
 net/mac80211/mesh_plink.c |    6 ++++++
 1 files changed, 6 insertions(+), 0 deletions(-)

diff --git a/net/mac80211/mesh_plink.c b/net/mac80211/mesh_plink.c
index f4124d7..6209327 100644
--- a/net/mac80211/mesh_plink.c
+++ b/net/mac80211/mesh_plink.c
@@ -302,6 +302,12 @@ static struct sta_info *mesh_peer_init(struct ieee80211_sub_if_data *sdata,
 	else
 		memset(&sta->sta.ht_cap, 0, sizeof(sta->sta.ht_cap));
 
+	if (elems->ht_operation)
+		if (!(elems->ht_operation->ht_param &
+		      IEEE80211_HT_PARAM_CHAN_WIDTH_ANY))
+			sta->sta.ht_cap.cap &=
+					    ~IEEE80211_HT_CAP_SUP_WIDTH_20_40;
+
 	rate_control_rate_init(sta);
 	spin_unlock_bh(&sta->lock);
 
-- 
1.7.5.4

Re: [PATCH 2/2] mac80211: don't transmit 40MHz frames to 20MHz peer

[Johannes Berg]

On Thu, 2012-04-26 at 15:01 -0700, Thomas Pedersen wrote:
> If a mesh peer indicates it is operating as 20MHz-only in its HT
> operation IE, have the rate control algorithm respect this by disabling
> the equivalent bit in the ieee80211_sta HT capabilities.
> 
> Signed-off-by: Thomas Pedersen <thomas@cozybit.com>
> ---
>  net/mac80211/mesh_plink.c |    6 ++++++
>  1 files changed, 6 insertions(+), 0 deletions(-)
> 
> diff --git a/net/mac80211/mesh_plink.c b/net/mac80211/mesh_plink.c
> index f4124d7..6209327 100644
> --- a/net/mac80211/mesh_plink.c
> +++ b/net/mac80211/mesh_plink.c
> @@ -302,6 +302,12 @@ static struct sta_info *mesh_peer_init(struct ieee80211_sub_if_data *sdata,
>  	else
>  		memset(&sta->sta.ht_cap, 0, sizeof(sta->sta.ht_cap));
>  
> +	if (elems->ht_operation)
> +		if (!(elems->ht_operation->ht_param &
> +		      IEEE80211_HT_PARAM_CHAN_WIDTH_ANY))
> +			sta->sta.ht_cap.cap &=
> +					    ~IEEE80211_HT_CAP_SUP_WIDTH_20_40;

The double if is a bit odd, why not use && to clean up all the
formatting?

johannes

Re: [PATCH 2/2] mac80211: don't transmit 40MHz frames to 20MHz peer

[Thomas Pedersen]

On Thu, May 3, 2012 at 11:55 AM, Johannes Berg
<johannes@sipsolutions.net> wrote:
> On Thu, 2012-04-26 at 15:01 -0700, Thomas Pedersen wrote:
>> If a mesh peer indicates it is operating as 20MHz-only in its HT
>> operation IE, have the rate control algorithm respect this by disabling
>> the equivalent bit in the ieee80211_sta HT capabilities.
>>
>> Signed-off-by: Thomas Pedersen <thomas@cozybit.com>
>> ---
>>  net/mac80211/mesh_plink.c |    6 ++++++
>>  1 files changed, 6 insertions(+), 0 deletions(-)
>>
>> diff --git a/net/mac80211/mesh_plink.c b/net/mac80211/mesh_plink.c
>> index f4124d7..6209327 100644
>> --- a/net/mac80211/mesh_plink.c
>> +++ b/net/mac80211/mesh_plink.c
>> @@ -302,6 +302,12 @@ static struct sta_info *mesh_peer_init(struct ieee80211_sub_if_data *sdata,
>>       else
>>               memset(&sta->sta.ht_cap, 0, sizeof(sta->sta.ht_cap));
>>
>> +     if (elems->ht_operation)
>> +             if (!(elems->ht_operation->ht_param &
>> +                   IEEE80211_HT_PARAM_CHAN_WIDTH_ANY))
>> +                     sta->sta.ht_cap.cap &=
>> +                                         ~IEEE80211_HT_CAP_SUP_WIDTH_20_40;
>
> The double if is a bit odd, why not use && to clean up all the
> formatting?

Good point, but later Ashok adds a patch which takes the same branch anyway.

Thomas