Johannes Berg <johannes@xxxxxxxxxxxxxxxx> wrote:
>
> We had this precise discussion previously for scanning, and just like
> there, fixed in ab9177d83c04 ("wifi: mac80211: don't use rate mask for
> scanning"), I feel the right way to approach this issue here would be to
> similarly not use the rate mask for offchannel TX, which is I think
> pretty much the same situation, you could have a rate mask set for only
> 2.4 GHz where the connection is (and empty for other bands), which is
> accepted by cfg80211 and mac80211, but then do offchannel TX on 5 GHz
> anyway.
>
> So I think the right way to approach this would be to do something like
>
> diff --git a/net/mac80211/offchannel.c b/net/mac80211/offchannel.c
> index 28d03196ef75..33361b4d9acf 100644
> --- a/net/mac80211/offchannel.c
> +++ b/net/mac80211/offchannel.c
> @@ -830,6 +830,8 @@ int ieee80211_mgmt_tx(struct wiphy *wiphy, struct wireless_dev *wdev,
> flags = IEEE80211_TX_INTFL_NL80211_FRAME_TX |
> IEEE80211_TX_CTL_REQ_TX_STATUS;
>
> + flags |= IEEE80211_TX_CTRL_SCAN_TX;
> +
> if (params->no_cck)
> flags |= IEEE80211_TX_CTL_NO_CCK_RATE;
>
>
> though at that point we need to rename that flag too, I guess.
Thanks for the suggestions. I made and sent a patch [1] that works well in
my side using syzbot's reproducer code.
[1] https://lore.kernel.org/linux-wireless/20240729074816.20323-1-pkshih@xxxxxxxxxxx/T/#u
