Received-SPF: pass (google.com: domain of linux-wireless+bounces-8753-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:40f1:3f00::1 as permitted sender) client-ip=2604:1380:40f1:3f00::1;
From: Kalle Valo <kvalo@kernel.org>
To: Rameshkumar Sundaram <quic_ramess@quicinc.com>
Cc: <ath12k@lists.infradead.org>,  <linux-wireless@vger.kernel.org>
Subject: Re: [PATCH] wifi: ath12k: modify remain on channel for single wiphy
References: <20240528082739.1226758-1-quic_ramess@quicinc.com>
Date: Mon, 10 Jun 2024 16:56:11 +0300
In-Reply-To: <20240528082739.1226758-1-quic_ramess@quicinc.com> (Rameshkumar
	Sundaram's message of "Tue, 28 May 2024 13:57:39 +0530")
Message-ID: <878qzcq4uc.fsf@kernel.org>
User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/28.2 (gnu/linux)
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain

Rameshkumar Sundaram <quic_ramess@quicinc.com> writes:

> When multiple radios are advertised as a single wiphy which
> supports various bands, vdev creation for the vif is deferred
> until channel is assigned to it.
> If a remain on channel(RoC) request is received from mac80211,
> select the corresponding radio(ar) based on channel and create
> a vdev on that radio to initiate an RoC scan.
>
> Note that on RoC completion this vdev is not deleted. If a new
> RoC/hw scan request is seen on that same vif for a different band the
> vdev will be deleted and created on the new radio supporting the
> request.
>
> Also if the RoC scan is requested when the vdev is in started state,
> no switching to new radio is allowed and RoC request can be accepted
> only on channels within same radio.
>
> Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
> Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0.c5-00481-QCAHMTSWPL_V1.0_V2.0_SILICONZ-3
>
> Signed-off-by: Rameshkumar Sundaram <quic_ramess@quicinc.com>

I did some white space changes to the commit message.

> @@ -8416,12 +8416,63 @@ static int ath12k_mac_op_remain_on_channel(struct ieee80211_hw *hw,
>  	struct ath12k_vif *arvif = ath12k_vif_to_arvif(vif);
>  	struct ath12k_hw *ah = ath12k_hw_to_ah(hw);
>  	struct ath12k_wmi_scan_req_arg arg;
> -	struct ath12k *ar;
> +	struct ath12k *ar, *prev_ar;
>  	u32 scan_time_msec;
> +	bool create = true;
>  	int ret;
>  
> -	ar = ath12k_ah_to_ar(ah, 0);
> +	if (ah->num_radio == 1) {
> +		WARN_ON(!arvif->is_created);
> +		ar = ath12k_ah_to_ar(ah, 0);
> +		goto scan;
> +	}
> +
> +	ar = ath12k_mac_select_scan_device(hw, vif, chan->center_freq);
> +	if (!ar)
> +		return -EINVAL;
> +
> +	/* If the vif is already assigned to a specific vdev of an ar,
> +	 * check whether its already started, vdev which is started
> +	 * are not allowed to switch to a new radio.
> +	 * If the vdev is not started, but was earlier created on a
> +	 * different ar, delete that vdev and create a new one. We don't
> +	 * delete at the scan stop as an optimization to avoid redundant
> +	 * delete-create vdev's for the same ar, in case the request is
> +	 * always on the same band for the vif
> +	 */
> +	if (arvif->is_created) {
> +		if (WARN_ON(!arvif->ar))
> +			return -EINVAL;
> +
> +		if (ar != arvif->ar && arvif->is_started)
> +			return -EINVAL;

I wonder if -EBUSY would be more descriptive here? I changed to that in
the pending branch.

> +		if (ar != arvif->ar) {
> +			/* backup the previously used ar ptr, since the vdev delete
> +			 * would assign the arvif->ar to NULL after the call
> +			 */
> +			prev_ar = arvif->ar;
> +			mutex_lock(&prev_ar->conf_mutex);
> +			ret = ath12k_mac_vdev_delete(prev_ar, vif);
> +			mutex_unlock(&prev_ar->conf_mutex);
> +			if (ret)
> +				ath12k_warn(prev_ar->ab,
> +					    "unable to delete scan vdev %d\n", ret);

Do we really want to continue if vdev_delete() fails? In the pending
branch I added 'return ret' here and modified the warning message a bit.

> +		} else {
> +			create = false;
> +		}
> +	}
> +
> +	if (create) {
> +		mutex_lock(&ar->conf_mutex);
> +		ret = ath12k_mac_vdev_create(ar, vif);
> +		mutex_unlock(&ar->conf_mutex);
> +		if (ret) {
> +			ath12k_warn(ar->ab, "unable to create scan vdev %d\n", ret);
> +			return -EINVAL;
> +		}

Also here I modified the warning message a bit.

The pending commit here:

https://git.kernel.org/pub/scm/linux/kernel/git/kvalo/ath.git/commit/?h=pending&id=9b4ec32e921b34bd7a03d39cc0a75cba7e85dc02

-- 
https://patchwork.kernel.org/project/linux-wireless/list/

https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches