Hi, I suggest here some improvements to this patch but I don't have strong opinion about them so any feedback is appreciated:
-split it into two: one for storing the current vif into brcms_c_info and also setting it to NULL on remove_interface to avoid dangling pointers; and an other one for actually updating the beacon; I believe them to be separate changes. On Sun, 23 Sep 2018 12:54:25 +0300 Ali MJ Al-Nasrawy <[email protected]> wrote: > +static int brcms_ops_beacon_set_tim(struct ieee80211_hw *hw, > + struct ieee80211_sta *sta, bool set) > +{ > + struct brcms_info *wl = hw->priv; > + struct sk_buff *beacon; > + u16 tim_offset = 0; > + > + beacon = ieee80211_beacon_get_tim(hw, wl->wlc->vif, > + &tim_offset, NULL); > + if (beacon){ > + spin_lock_bh(&wl->lock); > + brcms_c_set_new_beacon(wl->wlc, beacon, tim_offset, > + > wl->wlc->vif->bss_conf.dtim_period); > + spin_unlock_bh(&wl->lock); > + } > + > + return 0; > +} > + -on brcms_ops_beacon_set_tim, include ieee80211_beacon_get_tim call in the lock context and also check for vif to be null before passing it as argument.
