Re: [PATCH wireless 1/2] wifi: rtw88: program the channel when the device is started

[Luka Gejak]

On Wed Oct 7, 2026 at 11:12 AM CEST, Jeremy Fareau wrote:
> The chip is reset by power_off()/power_on(), but mac80211 only calls
> ieee80211_ops::config() when its own channel state changes.  After a
> stop/start cycle, such as
>
>     ip link set <dev> down
>     ip link set <dev> up
>

I don't think there is need to say what stop/start cycle is in commit
message.

> the channel is unchanged from mac80211's point of view, so the radio is
> never reprogrammed.  The device then listens on whatever channel the
> hardware came up on while iw reports the configured one, and receives
> almost nothing.
>
> Observed with an RTL8814AU (ALFA AWUS1900) in monitor mode on an
> aarch64 host.  A 25 s capture on channel 6 after a down/up cycle yields
> 3 frames, and 431 frames as soon as any real channel change is
> requested.  An RTL8812AU (Linksys WUSB6300) on the same host, the same
> channel and the same second yields 2438 frames.
>
> Program the channel in rtw_ops_start(), like rtw_ips_pwr_up() already
> does when leaving IPS.
>
> Fixes: e3037485c68e ("rtw88: new Realtek 802.11ac driver")
> Assisted-by: LLM
> Signed-off-by: Jeremy Fareau <jeremy.fareau@xxxxxxxxx>
> ---
>  drivers/net/wireless/realtek/rtw88/mac80211.c | 11 +++++++++++
>  1 file changed, 11 insertions(+)
>
> diff --git a/drivers/net/wireless/realtek/rtw88/mac80211.c b/drivers/net/wireless/realtek/rtw88/mac80211.c
> index b01b98d24b0a..827f38390969 100644
> --- a/drivers/net/wireless/realtek/rtw88/mac80211.c
> +++ b/drivers/net/wireless/realtek/rtw88/mac80211.c
> @@ -57,6 +57,17 @@ static int rtw_ops_start(struct ieee80211_hw *hw)
>  
>  	mutex_lock(&rtwdev->mutex);
>  	ret = rtw_core_start(rtwdev);
> +
> +	/* The chip is reset by power_off()/power_on(), but mac80211 only calls
> +	 * ieee80211_ops::config() when its own channel state changes. After a
> +	 * stop/start cycle the channel is unchanged from mac80211's point of
> +	 * view, so the radio would never be reprogrammed and the device would
> +	 * receive nothing. Program it here, like rtw_ips_pwr_up() already does
> +	 * when leaving IPS.
> +	 */

We don't use that comment style anymore. Change it to:

/*
 * text
 * text
 */

Also that comment is rather extensive, maybe shorten it to something
like:

/*
 * The chip is reset by power_off() and power_on(), but mac80211 won't
 * call config() after stop/start since its channel state is unchanged.
 * Reprogram the channel here, as rtw_ips_pwr_up() does.
 */

> +	if (!ret && hw->conf.chandef.chan)
> +		rtw_set_channel(rtwdev);
> +
>  	mutex_unlock(&rtwdev->mutex);
>  
>  	return ret;

Besides that I noticed that there are many double spaces after
punctuation in commit message, so please fix that too. And
should this cc stable?

Best regards,
Luka Gejak