Received-SPF: pass (google.com: domain of linux-wireless-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Sender: tmariyap=codeaurora.org@mg.codeaurora.org
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8;
 format=flowed
Content-Transfer-Encoding: 8bit
Date:   Thu, 06 May 2021 23:19:06 +0530
From:   Thiraviyam Mariyappan <tmariyap@codeaurora.org>
To:     Johannes Berg <johannes@sipsolutions.net>
Cc:     ath11k@lists.infradead.org, linux-wireless@vger.kernel.org
Subject: Re: [PATCHv2] mac80211: increment rx stats according to USES_RSS flag
In-Reply-To: <ec30381c062e3eb87abb724641a15331cfc3d11c.camel@sipsolutions.net>
References: <1613563010-1489-1-git-send-email-tmariyap@codeaurora.org>
 (sfid-20210217_125904_154301_738B3086)
 <c0aef41d2ecf09188de372fe4f7d6b1954e54e07.camel@sipsolutions.net>
 <1ee8d562986128767c037d20aedb96a5@codeaurora.org>
 <ec30381c062e3eb87abb724641a15331cfc3d11c.camel@sipsolutions.net>
Message-ID: <df6e33fe198e179f1c1f31ca6d55e995@codeaurora.org>
User-Agent: Roundcube Webmail/1.3.9
Precedence: bulk

On 2021-04-23 13:28, Johannes Berg wrote:
> On Wed, 2021-04-21 at 22:18 +0530, Thiraviyam Mariyappan wrote:
>> In case of Mesh fast_rx is not applicable, but still USES_RSS can be
>> enabled from driver when parallel RX is supported by HW/Driver,
>> right? 
> 
> Yes, I guess that's true.
> 
>> Hence checked for USES_RSS support to update per cpu stats.Please
>> correct me if the meaning of USES_RSS is misunderstood and it applies
>> only when fast_rx for a STA is enabled.
>> 
> 
> Well, actually using multi-queue is pointless or even 
> counter-productive
> when you don't have fast-RX, since then you'll run into a common lock,
> and doing much processing on multiple CPUs but under a common lock 
> might
> well be worse than doing it on a single CPU in the first place, since
> you'll bounce the lock around all the time.
> 
> However, you're right that the driver might generally advertise
> USES_RSS, but then not do it for mesh, but that throws off some
> statistics.
> 
> Something like this might then be a much better fix though?
Below fix good to me and working fine.
> 
> 
> diff --git a/net/mac80211/sta_info.c b/net/mac80211/sta_info.c
> index ec6973ee88ef..f87e883862d9 100644
> --- a/net/mac80211/sta_info.c
> +++ b/net/mac80211/sta_info.c
> @@ -2092,7 +2092,7 @@ sta_get_last_rx_stats(struct sta_info *sta)
>  	struct ieee80211_local *local = sta->local;
>  	int cpu;
> 
> -	if (!ieee80211_hw_check(&local->hw, USES_RSS))
> +	if (!sta->pcpu_rx_stats)
>  		return stats;
> 
>  	for_each_possible_cpu(cpu) {
> @@ -2192,9 +2192,7 @@ static void sta_set_tidstats(struct sta_info 
> *sta,
>  	int cpu;
> 
>  	if (!(tidstats->filled & BIT(NL80211_TID_STATS_RX_MSDU))) {
> -		if (!ieee80211_hw_check(&local->hw, USES_RSS))
> -			tidstats->rx_msdu +=
> -				sta_get_tidstats_msdu(&sta->rx_stats, tid);
> +		tidstats->rx_msdu += sta_get_tidstats_msdu(&sta->rx_stats, tid);
> 
>  		if (sta->pcpu_rx_stats) {
>  			for_each_possible_cpu(cpu) {
> @@ -2308,8 +2306,7 @@ void sta_set_sinfo(struct sta_info *sta, struct
> station_info *sinfo,
> 
>  	if (!(sinfo->filled & (BIT_ULL(NL80211_STA_INFO_RX_BYTES64) |
>  			       BIT_ULL(NL80211_STA_INFO_RX_BYTES)))) {
> -		if (!ieee80211_hw_check(&local->hw, USES_RSS))
> -			sinfo->rx_bytes += sta_get_stats_bytes(&sta->rx_stats);
> +		sinfo->rx_bytes += sta_get_stats_bytes(&sta->rx_stats);
> 
>  		if (sta->pcpu_rx_stats) {
>  			for_each_possible_cpu(cpu) {
> 
> 
> johannes